THE CANARIAS INFRARED CAMERA EXPERIMENT (CIRCE)...
Transcript of THE CANARIAS INFRARED CAMERA EXPERIMENT (CIRCE)...
THE CANARIAS INFRARED CAMERA EXPERIMENT (CIRCE) AND THE SEARCHFOR AND STUDY OF MASSIVE STARS
By
MICHELLE L. EDWARDS
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOLOF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
2008
1
c© 2008 Michelle L. Edwards
2
To my husband, David.
Now I feel no rain, no cold, no loneliness – for you are my shelter, my warmth, and mycompanion. We are two people but there is one life before us. May beauty surround us in
the journey ahead and through all the years to come.
3
ACKNOWLEDGMENTS
This dissertation is the result of the help, talent, and encouragement of many people
whom I would like to thank. First, I acknowledge my advisers, Drs. Stephen Eikenberry
and Reba Bandyopadhyay. Over the last 2 years, Reba has offered impeccable advice
about my science, writing, and career as a scientist and I look to her as a role model and
mentor. I feel secure in saying that this work would not exist in its current state if she had
not agreed to be my co-adviser.
While there is little I can say or do to repay Steve for his friendship, advice, and
constant cheerleading during my early graduate career, what he gave me in these last
years was the most valuable gift an adviser can give a senior student - a push out of the
nest and the freedom to make and follow my own path. I am certain I would not have
developed the self-confidence and poise necessary to write this dissertation and defend it
without that experience.
I thank my committee, Dr. Charlie Telesco, Dr. Ata Sarajedini, Dr. Guido Mueller,
and Dr. Rafael Guzman for useful input and discussions. I also acknowledge Dr. Fred
Hamann for many chats and a great introduction to stellar winds.
I thank P.E.O. International for awarding me a P.E.O. Scholar Award, which helped
support me financially during the last two years of my graduate career. I am especially
grateful to Anne Seymour and Chapter H in Baltimore, MD for nominating me.
I acknowledge many close friends whose help and camaraderie made the last seven
years of graduate school navigable. These include Rachel and Marcus Chen, Dan and
Jodi Bell, Veronica Donoso, Joe Carson, Andrew Gosling, Mike Barker, Ashley Espy,
Valerie Mikles, Ana Matkovic, Suvrath Mahadevan, Maren Hempel, Ileana Vass, Veera
Boonyasait, Joanna Levine, Eric McKenzie, Dan Capellupo, Ineta Jonusas, Deborah
Herbstman, Dimitri Veras and especially, my classmate, constant companion, and dearest
friend, David Clark.
4
I also acknowledge my CIRCE teammates, Miguel Charcos-Llorens, Nestor Lasso, and
Dr. Antonio Marin-Franch, as well as the UF instrumentation group, particularly, Dr. S.
Nicholas Raines, Dr. Chris Packham, and Jeff Julian, with whom it was a pleasure and
privilege to work. I thank Dr. Craig Warner whose tireless efforts on the FATBOY data
reduction pipeline made Chapters 5 and 6 possible.
Throughout my life, I have been lucky to have excellent teachers who inspired and
encouraged me. To a few, I owe many, many thanks: Dr. John Oliver, Dr. Robert Boyle,
Drs. Ken and Priscilla Laws, Mr. Joe Coleman, Mrs. Karen Fishman, Mr. Neil Wood,
and especially Dr. Tony Morgan, my close friend and adviser, who took me to Flagstaff,
pointed me in the direction of a telescope, and changed my life forever.
I acknowledge the many gifted choirs and choir directors in Gainesville who welcomed
me into their folds, especially Dr. William Kesling and the UF University Choir; it was a
great honor to sing with such consummate professionals.
I would like to acknowledge my loving family including my mother, Linda, who
pursued her own college degree while raising two small children and inspired me to be a
fighter and never give up on my dreams, my father Walter, a constant source of support
and encouragement whose passion for history showed me what it was to really love to
study something, and my hilarious, musically-talented brother Matthew and his smart,
beautiful wife Jackie who have brought incredible joy and many laughs into my life. I also
thank the entire Carney and Edwards families for cheering me on through years and years
(and years) of school.
I thank my parents-in-law, Marie and Paul Edmeades, kindred spirits, whose support
and love throughout the years (and especially in the last few months) has been a source
of constant joy in our lives. I thank their many friends and extended family members who
have also supported me.
Last but not least, I acknowledge my husband and love for 9 years, David Edmeades,
to whom I dedicate this work. Wherever you are, dear, I call home.
5
TABLE OF CONTENTS
page
ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
CHAPTER
1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.1 The Gran Telescopio Canarias . . . . . . . . . . . . . . . . . . . . . . . . . 181.2 A NIR Visitor Instrument . . . . . . . . . . . . . . . . . . . . . . . . . . . 191.3 Science with CIRCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201.4 The Search for and Study of Massive Stars . . . . . . . . . . . . . . . . . . 221.5 Observations of Magnetars . . . . . . . . . . . . . . . . . . . . . . . . . . . 231.6 Environments of Soft Gamma Repeaters (SGRs) . . . . . . . . . . . . . . . 251.7 The Locations of Luminous Blue Variables (LBVs) . . . . . . . . . . . . . 28
2 OPTICAL DESIGN AND ANALYSIS OF CIRCE . . . . . . . . . . . . . . . . . 33
2.1 Preliminary Optical Designs . . . . . . . . . . . . . . . . . . . . . . . . . . 332.2 Introduction to Optical Design Software and Terminology . . . . . . . . . . 362.3 Envelope Re-design and GTC Integration . . . . . . . . . . . . . . . . . . . 392.4 Final Optical Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412.5 Optical Analysis of Imaging Mode . . . . . . . . . . . . . . . . . . . . . . . 412.6 Spectroscopy and Polarimetry . . . . . . . . . . . . . . . . . . . . . . . . . 442.7 Tolerancing Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3 OPTO-MECHANICAL DESIGN OF CIRCE . . . . . . . . . . . . . . . . . . . 64
3.1 Basics of Mechanical Design . . . . . . . . . . . . . . . . . . . . . . . . . . 643.2 Mechanical Design Software . . . . . . . . . . . . . . . . . . . . . . . . . . 653.3 Designing and Drawing the CIRCE Mirrors . . . . . . . . . . . . . . . . . 673.4 Opto-Mechanical Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . 713.5 Two-Dimensional Opto-Mechanical Layout . . . . . . . . . . . . . . . . . . 723.6 Design and Analysis of Three-Dimensional Mechanical Layout . . . . . . . 743.7 Development in the Manufacture of the CIRCE Optics . . . . . . . . . . . 753.8 Lightweighting the CIRCE Optics . . . . . . . . . . . . . . . . . . . . . . . 763.9 Finalized Mechanical Drawings of the CIRCE Mirrors . . . . . . . . . . . . 783.10 Mechanical Design of the Mirror Brackets . . . . . . . . . . . . . . . . . . 783.11 Preliminary Designs of the Optical Bench . . . . . . . . . . . . . . . . . . . 81
6
4 CRYO-MECHANICAL DESIGN OF THE CIRCE PUPIL BOX . . . . . . . . . 109
4.1 Introduction to Cryo-Mechanisms . . . . . . . . . . . . . . . . . . . . . . . 1094.2 Optical Elements in the Pupil Box . . . . . . . . . . . . . . . . . . . . . . 1104.3 Design of the Pupil Box Components . . . . . . . . . . . . . . . . . . . . . 111
4.3.1 The Filter, Grism, and Lyot Cartridges . . . . . . . . . . . . . . . . 1114.3.2 The Filter, Grism, and Lyot Wheels . . . . . . . . . . . . . . . . . . 1124.3.3 Wheel Hubs and Bearing Races . . . . . . . . . . . . . . . . . . . . 1134.3.4 Filter, Grism, and Lyot Gears . . . . . . . . . . . . . . . . . . . . . 1134.3.5 Design of the Pupil Box . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.4 Integration of the Pupil Box Mechanism . . . . . . . . . . . . . . . . . . . 115
5 THE SEARCH FOR MASSIVE STARS AROUND SGR 1806-20 . . . . . . . . 136
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1365.2 Narrow-Band Photometry . . . . . . . . . . . . . . . . . . . . . . . . . . . 1375.3 Observations and Data Reduction . . . . . . . . . . . . . . . . . . . . . . . 1405.4 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.4.1 Photometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1415.4.2 Star Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1445.4.3 Limiting Magnitudes and Match Drop-Outs . . . . . . . . . . . . . . 1455.4.4 Field Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1505.4.5 Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1515.4.6 Calculating Equivalent Widths . . . . . . . . . . . . . . . . . . . . . 152
5.5 Comparison to Previous Results . . . . . . . . . . . . . . . . . . . . . . . . 1535.5.1 Wolf Rayet Stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1545.5.2 Luminous Blue Variable (LBV) 1806-20 . . . . . . . . . . . . . . . . 156
5.5.2.1 Evaluation of PSF photometry . . . . . . . . . . . . . . . 1565.5.2.2 Spectroscopy of LBV 1806-20 . . . . . . . . . . . . . . . . 157
5.5.3 OBI stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1585.6 New Results on Observed Stellar Populations in My Field . . . . . . . . . 160
5.6.1 New Candidate Cluster Members . . . . . . . . . . . . . . . . . . . 1605.6.2 Foreground Stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6 PROBING MAGNETAR ENVIRONMENTS . . . . . . . . . . . . . . . . . . . 174
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1746.1.1 SGR 1900+14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1746.1.2 SGR 1627-41 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1756.1.3 SGR 0526-66 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
6.2 Application of the Narrow-band Imaging Technique . . . . . . . . . . . . . 1766.3 Observations and Data Reduction . . . . . . . . . . . . . . . . . . . . . . . 1776.4 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
6.4.1 Analysis of 1900+14 . . . . . . . . . . . . . . . . . . . . . . . . . . . 1796.4.1.1 Photometry . . . . . . . . . . . . . . . . . . . . . . . . . . 179
7
6.4.1.2 Star matching . . . . . . . . . . . . . . . . . . . . . . . . . 1806.4.1.3 Limiting magnitudes and match drop-outs . . . . . . . . . 1806.4.1.4 Field selection . . . . . . . . . . . . . . . . . . . . . . . . . 1826.4.1.5 Calculating errors and equivalent widths . . . . . . . . . . 183
6.4.2 Analysis of 1627-41 . . . . . . . . . . . . . . . . . . . . . . . . . . . 1846.4.2.1 Departures from the standard technique . . . . . . . . . . 1846.4.2.2 Photometry and band matching . . . . . . . . . . . . . . . 1856.4.2.3 Limiting magnitudes and match drop-outs . . . . . . . . . 1866.4.2.4 Photometric errors and equivalent width calculations . . . 187
6.4.3 Analysis of 0526-66 . . . . . . . . . . . . . . . . . . . . . . . . . . . 1886.4.3.1 Photometry and band matching . . . . . . . . . . . . . . . 1886.4.3.2 Limiting magnitudes and match drop-outs . . . . . . . . . 1896.4.3.3 Field selection, errors, and equivalent widths . . . . . . . . 190
6.5 Preliminary Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1916.5.1 SGR 1900+14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1916.5.2 SGR 1627-41 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1936.5.3 SGR 0526-66 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
6.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
7 LUMINOUS BLUE VARIABLE IN THEIR HOST CLUSTERS . . . . . . . . . 217
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2177.2 Sample Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2187.3 Probability Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
7.3.1 Individual Analyses of LBVs as Peripheral Cluster Members . . . . 2237.3.2 Ensemble Analysis of LBVs as Peripheral Cluster Members . . . . . 224
7.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2267.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
8 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
8.1 Summary of the CIRCE Project . . . . . . . . . . . . . . . . . . . . . . . . 2328.2 Status of CIRCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2348.3 Summary of A NIR Narrow-Band Imaging Survey . . . . . . . . . . . . . . 2358.4 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
8.4.1 Extinction and Stellar Populations in the SGR Fields . . . . . . . . 2368.4.2 Massive Stars with CIRCE . . . . . . . . . . . . . . . . . . . . . . . 2378.4.3 Magnetar Environments . . . . . . . . . . . . . . . . . . . . . . . . 237
APPENDIX: ZEMAX OUTPUT OF CIRCE PRESCRIPTION . . . . . . . . . . . . 239
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
BIOGRAPHICAL SKETCH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
8
LIST OF TABLES
Table page
2-1 Results from initial tolerancing run . . . . . . . . . . . . . . . . . . . . . . . . . 48
2-2 Results from the final tolerancing run . . . . . . . . . . . . . . . . . . . . . . . . 49
2-3 Final tolerances for decenters and tilts . . . . . . . . . . . . . . . . . . . . . . . 50
3-1 Dimensions of the CIRCE mirrors . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5-1 Summary of previous work on Cl 1806-20 . . . . . . . . . . . . . . . . . . . . . . 164
5-2 Properties of known massive stars in Cl 1806-20 . . . . . . . . . . . . . . . . . . 165
5-3 Select spectrial lines in LBV 1806-20 . . . . . . . . . . . . . . . . . . . . . . . . 165
5-4 Properties of candidate OBI stars in Cl 1806-20 . . . . . . . . . . . . . . . . . . 166
6-1 Properties of known massive stars near SGR 1900+14 . . . . . . . . . . . . . . . 198
6-2 Properties of candidate massive stars near SGR 1900+14 . . . . . . . . . . . . . 199
6-3 Properties of candidate massive stars near SGR 1627-41 found using Brγ . . . . 200
6-4 Properties of candidate massive stars near SGR 1627-41 found using HeI . . . . 201
7-1 Positions of LBVs in host clusters . . . . . . . . . . . . . . . . . . . . . . . . . . 230
7-2 Cluster mass enclosed by LBV radius . . . . . . . . . . . . . . . . . . . . . . . . 230
A-1 ZEMAX surface data summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
A-2 ZEMAX coordinate break summary . . . . . . . . . . . . . . . . . . . . . . . . . 241
9
LIST OF FIGURES
Figure page
1-1 Gran Telescopio Canarias mirrors and foci . . . . . . . . . . . . . . . . . . . . . 29
1-2 Giant SGR burst . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1-3 Soft gamma-ray flare in a magnetar . . . . . . . . . . . . . . . . . . . . . . . . . 31
1-4 Spin period vs. period derivative . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2-1 Initial refractive design of CIRCE . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2-2 Preliminary all-reflective design of CIRCE . . . . . . . . . . . . . . . . . . . . . 52
2-3 Standard coordinate system for optical design software . . . . . . . . . . . . . . 53
2-4 ZEMAX user interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2-5 Radius of curvature and parent mirrors . . . . . . . . . . . . . . . . . . . . . . . 55
2-6 Layout showing CIRCE parent mirrors . . . . . . . . . . . . . . . . . . . . . . . 56
2-7 Folded Cassegrain envelope of the GTC . . . . . . . . . . . . . . . . . . . . . . . 57
2-8 Envelope re-design for CIRCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2-9 Final optical design of CIRCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2-10 Footprint diagram at detector focal plane . . . . . . . . . . . . . . . . . . . . . . 60
2-11 J-band encircled energy diagram . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2-12 Encircled energy diagram for the K-band . . . . . . . . . . . . . . . . . . . . . . 62
2-13 Spot diagram for the K-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3-1 Representation of mirrors in ZEMAX vs. AutoCAD . . . . . . . . . . . . . . . . 84
3-2 Cross-section of a simple mirror . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3-3 Footprint diagrams of the CIRCE mirrors . . . . . . . . . . . . . . . . . . . . . 86
3-4 Preliminary XY cross-section of Collimator 1 . . . . . . . . . . . . . . . . . . . . 87
3-5 Preliminary YZ cross-section of Collimator 1 . . . . . . . . . . . . . . . . . . . . 88
3-6 Preliminary two-dimensional mechanical drawing of Collimator 1 . . . . . . . . 89
3-7 Revised two-dimensional drawing of Collimator 1 . . . . . . . . . . . . . . . . . 90
3-8 Original designs of fold mirror . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
10
3-9 Revised fold mirror . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3-10 Two-dimensional opto-mechanical layout . . . . . . . . . . . . . . . . . . . . . . 93
3-11 Solid model of Imager 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3-12 Solid model of CIRCE mirrors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3-13 Image of the CIRCE practice mirrors . . . . . . . . . . . . . . . . . . . . . . . . 96
3-14 Cross-section of a lightweighting pattern . . . . . . . . . . . . . . . . . . . . . . 97
3-15 Solid model of a lightweight CIRCE optic . . . . . . . . . . . . . . . . . . . . . 98
3-16 Final mechanical drawing for Collimator 1 blank . . . . . . . . . . . . . . . . . . 99
3-17 Final optical drawing for Collimator 1 . . . . . . . . . . . . . . . . . . . . . . . 100
3-18 Examples of L and T brackets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3-19 Redesigned bracket for Imager 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3-20 Bracket for Imager 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
3-21 Compound bracket for Imager 4 and Collimator 1 . . . . . . . . . . . . . . . . . 104
3-22 Mechanical drawings of a CIRCE bracket . . . . . . . . . . . . . . . . . . . . . . 105
3-23 Completed CIRCE bracket . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3-24 Preliminary solid model of the lightweight CIRCE bench . . . . . . . . . . . . . 107
3-25 Final solid model of the CIRCE opto-mechanical layout . . . . . . . . . . . . . . 108
4-1 Example of a completed pupil box . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4-2 Example of a filter wheel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4-3 Optical layout of CIRCE with pupil box region . . . . . . . . . . . . . . . . . . 119
4-4 Broad-band filter for CIRCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4-5 Preliminary and final designs of a filter cartridge . . . . . . . . . . . . . . . . . 121
4-6 Solid model of a pupil wheel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4-7 Mechanical drawings of a pupil wheel . . . . . . . . . . . . . . . . . . . . . . . . 123
4-8 Mock-up of a complete pupil box . . . . . . . . . . . . . . . . . . . . . . . . . . 124
4-9 Solid model of a pupil wheel hub . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4-10 Solid model of a filter wheel and wheel hub . . . . . . . . . . . . . . . . . . . . . 126
11
4-11 Model of pupil box gears . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4-12 Solid model of a filter gear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4-13 Finished gear blank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4-14 Two-dimensional design of the pupil box . . . . . . . . . . . . . . . . . . . . . . 130
4-15 Solid model of the pupil box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4-16 Solid model of a motor mount . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
4-17 Solid model of pupil box and motor mounts . . . . . . . . . . . . . . . . . . . . 133
4-18 Solid model of a revised grism wheel . . . . . . . . . . . . . . . . . . . . . . . . 134
4-19 Solid model of the integrated filter wheel . . . . . . . . . . . . . . . . . . . . . . 135
5-1 Limiting J- and Ks- band magnitudes . . . . . . . . . . . . . . . . . . . . . . . . 167
5-2 Limiting magnitudes in the 3-band match . . . . . . . . . . . . . . . . . . . . . 168
5-3 Limiting magnitudes in the narrow-bands . . . . . . . . . . . . . . . . . . . . . 169
5-4 Initial color-color diagram of the field surrounding SGR 1806-20 . . . . . . . . . 170
5-5 Color-color diagram of the field surrounding SGR 1806-20 with magnitude cutoffs 171
5-6 Region of color-color diagram centered on cluster . . . . . . . . . . . . . . . . . 172
5-7 Spectra of LBV 1806-20s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
6-1 Limiting broad-band magnitudes for SGR 1900+14 . . . . . . . . . . . . . . . . 202
6-2 Initial color-color diagram for SGR 1900+14 . . . . . . . . . . . . . . . . . . . . 203
6-3 Narrow-band Brγ - K2.14 versus K2.14 diagram for SGR 1627-41 . . . . . . . . . 204
6-4 Narrow-band HeI - K2.03 versus K2.03 diagram for SGR 1627-41 . . . . . . . . . 205
6-5 Narrow-band Brγ - K2.14 versus K2.14 diagram for SGR 0526-66 . . . . . . . . . 206
6-6 Narrow-band HeI - K2.03 versus K2.03 diagram for SGR 0526-66 . . . . . . . . . 207
6-7 Color-color diagram of SGR 1900+14 with magnitude cutoffs . . . . . . . . . . . 208
6-8 Broad-band Ks-band image of SGR 1900+14 . . . . . . . . . . . . . . . . . . . . 209
6-9 Color-color diagram of SGR 1900+14 with new candidates . . . . . . . . . . . . 210
6-10 Narrow-band Brγ image of SGR 1627-41 . . . . . . . . . . . . . . . . . . . . . . 211
6-11 Narrow-band Brγ - K2.14 versus K2.14 diagram for SGR 0526-66 subfield 1 . . . 212
12
6-12 Narrow-band Brγ −K2.14 versus K2.14 diagram for SGR 0526-66 subfield 2 . . . 213
6-13 Narrow-band HeI - K2.03 versus K2.03 diagram for SGR 0526-66 subfield 1 . . . . 214
6-14 HeI - K2.03 versus K2.03 diagram for SGR 0526-66 subfield 2 . . . . . . . . . . . 215
6-15 Narrow-band Brγ image of SGR 0526-66 . . . . . . . . . . . . . . . . . . . . . . 216
7-1 Images of LBVs at the outskirts of host clusters . . . . . . . . . . . . . . . . . . 231
13
Abstract of Dissertation Presented to the Graduate Schoolof the University of Florida in Partial Fulfillment of theRequirements for the Degree of Doctor of Philosophy
THE CANARIAS INFRARED CAMERA EXPERIMENT (CIRCE) AND THE SEARCHFOR AND STUDY OF MASSIVE STARS
By
Michelle L. Edwards
August 2008
Chair: Stephen EikenberryCoChair: Reba BandyopadhyayMajor: Astronomy
I report on the design status of the Canarias InfraRed Camera Experiment (CIRCE),
an all-reflective near-infrared visitor instrument for the 10.4 meter Gran Telescopio
Canarias (GTC). In addition to functioning as a 1-2.5 micron imager, CIRCE will have
the capacity for narrow-band imaging, low- and moderate- resolution grism spectroscopy,
and imaging- and spectro- polarimetry. I present my contributions to the optical, opto-
and cryo- mechanical design.
I then detail my search for and study of massive stars around magnetars using the
current generation of near-infrared instruments. I survey the environments of Soft Gamma
Repeaters using near-infrared narrow-band imaging to probe for Brγ and HeI features
indicative of massive stars. Using this technique, I successfully detect previously known
massive stars in Cl 1806-20, the established natal cluster of SGR 1806-20. I also identify
new candidate massive stars in this cluster and around two other SGRs, which may
represent the bulk of the heretofore undiscovered massive star populations associated with
these objects.
Finally, I discuss a serendipitous discovery that developed from my study of massive
stars in clusters. Analyses of the five Luminous Blue Variables (LBVs) and LBV
candidates in massive cluster environments reveal that four of these rare, massive objects
lie on the outskirts of their host clusters. I demonstrate that the chance probability of
14
a cluster star lying consistently at or beyond the radii of these LBVs is 0.02%. Mass
segregation theories suggest that these stars, the most massive objects in their respective
clusters should fall inward over dynamical time-scales. Thus, while one would expect to
find a significant bias toward centrally located LBVs, my analyses show that these massive
stars are instead located on the periphery of their host clusters.
15
CHAPTER 1INTRODUCTION
In 1800, the famed astronomer William Herschel performed an experiment to measure
the temperatures of different colors of light. Using a prism, he dispersed sunlight into
a spectrum and placed a thermometer at the position of every color. As a control, he
also placed a thermometer slightly redward of the visible light. What he discovered
upon analyzing the results surprised him; the control thermometer was hotter than
the others. He postulated that it was warmed by invisible light from the sun, which he
called “calorific rays” (Herschel 1801). Today, we term the region of the electromagnetic
spectrum discovered by Herschel “infrared radiation” (Glass 1999).
Divided into three regimes, defined here as near- (1 - 5 µm), mid- (5 - 25 µm),
and far- (25 - 350 µm), infrared radiation is crucial to understanding a wide variety of
processes in our universe (Glass 1999). For example, near-infrared (NIR) observations
of the Galactic Plane have revealed millions of previously undetected stars, each whose
visible light is blocked by the dust and gas in the disk of our Galaxy (Skrutskie et al.
2006). Mid-infrared (MIR) studies of young stars have uncovered protoplanetary disks,
the precursors to exosolar planets and asteroid belts (Lagage & Pantin 1994; Telesco
2000). Meanwhile, cold dust clouds in the earliest phases of star formation (Mezger 1986;
Wszolek et al. 1989), are visible in the the far-infrared (FIR).
These discoveries and many others have only become possible in the last 40 years;
compared to the > 4000 year tradition of visible light observations (Chaisson & McMillan
1996), IR astronomy is a very young field. Successes in the early 1950s and 1960s with
lead-sulphide and gallium-doped germanium bolometers led to the first infrared sky
surveys, the Air Force Infrared Survey (Kleinmann et al. 1981) and the Two Micron Sky
Survey (Neugebauer & Leighton 1969). The 1970s gave rise to FIR balloon experiments
and the flight of the Kuiper Airborne Observatory, an infrared telescope housed in a
jet (Gillespie 1981; Low et al. 2007). Then in the 1980s, IR astronomy took a giant
16
leap forward with the development of IR arrays and the launch of IRAS, the InfraRed
Astronomical Satellite. The former heralded the coming of age of infrared instrumentation
and prepared the way for hundreds of modern infrared astronomical cameras (Rieke 2007).
The latter guaranteed continued interest in infrared science, unveiling > 350,000 infrared
sources.
Today, a variety of current and planned space-based infrared telescopes, including the
Spitzer Space Telescope, Herschel Space Observatory, and James Webb Space Telescope
(Werner et al. 2004; Pilbratt 2004; Sabelhaus et al. 2005), promise deep, high resolution
infrared images and spectra. Further, the location of large observatories above the water
vapor that absorbs infrared radiation, has made ground-based near- and mid- infrared
observing a reality. Every 8 - 10 meter class telescope now has a variety of infrared
instruments; most equipped with mercury cadmium telluride (HgCdTe) or indium
antimonide (InSb) detectors many times larger and more sensitive than the first arrays
(Rieke 2007). Additionally, a number of telescopes, like the 8-meter Gemini North and
South telescopes are themselves optimized for infrared science, utilizing special designs
and mirror coatings to limit emissivity, as well as undersized, lightweight secondary
mirrors capable of quick movements for MIR observing (Mountain et al. 1997; Roche
2004)
Following in the footsteps of great international observatories like Keck and Gemini
is the Gran Telescopio Canarias (GTC), the world’s largest optical/infrared telescope
(Alvarez et al. 2000; Rodriguez Espinosa & Alvarez Martin 2006). Also optimized for IR
observing, the GTC promises to be yet another positive step in the relatively new, but
clearly important field of IR astronomy. In the following sections, I will offer an overview
of the GTC and its collection of cutting-edge instruments. I will particularly focus on one
NIR instrument, the Canarias InfraRed Camera Experiment (CIRCE), the design of which
comprised a major part of my thesis work.
17
1.1 The Gran Telescopio Canarias
Located on the island of La Palma in the Canary Islands, the Gran Telescopio
Canarias (GTC) is a joint venture between Spain, Mexico, and the University of Florida
(UF). Currently in the final stages of construction, first science light is expected in late
2008. Upon its completion it will be the largest optical/infrared telescope in the world.
The GTC is a Ritchey-Chretien telescope, built with hyperbolic primary and
secondary mirrors to eliminate coma and spherical aberration. The primary is segmented
and composed of 36 hexagonal pieces each made of a specialized zero-expansion glass
ceramic manufactured by Schott Inc. The secondary is honeycomb-shaped and composed
of light-weight beryllium coated with nickel. The reduced mass of this mirror allows for
quick motions, or chopping, crucial for MIR observing (Rodrıguez Espinosa & Alvarez
2003; Rodriguez Espinosa & Alvarez Martin 2006). As of May 2008 the telescope structure
and dome are complete and in place at the Roque de los Muchachos Observatory. The
secondary mirror is mounted in its structure at the prime focus and 24 of the 36 primary
mirror segments are in place (Rodriguez-Espinosa, private communication). Once fully
assembled, the collecting surface of the primary mirror will be equivalent to a 10.4 meter
telescope.
Figure 1-1 shows the GTC’s mirrors and foci. The primary mirror is designated
M1, the secondary M2, and the movable tertiary M3. Light from the sky falling on the
primary is reflected to the secondary. When the tertiary is not in the optical path, light
from the secondary is reflected back down through a hole in the primary to the Cassegrain
Focus. When the tertiary is inserted into the GTC’s optical path it directs light from
the secondary perpendicular to the telescope’s native optical axis to one of four Bent or
Folded Cassegrain Foci (FCF) or one of two Nasymth Foci (NF). In all, the telescope can
accommodate 7 instruments, although astronomers can only observe with one at a time.
(Rodrıguez Espinosa & Alvarez 2003).
18
As with most modern large observatories, the GTC will have a suite of instruments
that cover the full range of optical and infrared wavelengths. While built by the GTC
partners, these cameras, often called facility instruments, belong to the observatory
and will be available to the entire community. Given the complexity and cost of
building cutting-edge facility instruments, the GTC has spread this process over
several generations. First generation instruments include CanariCam, a MIR imager,
spectrograph, polarimeter, and coronograph intended for the Cassegrain focus (Telesco
et al. 2003) and OSIRIS an optical spectrograph and imager that will mount at a
Nasmyth focus (Cepa et al. 2003). The GTC will commission both these instruments on
the telescope during the year after first scientific light. The observatory will commission
the second generation instruments EMIR, a NIR camera and high-resolution spectrograph
and FRIDA, a NIR integral field unit (IFU), upon their completion several years into the
telescope’s operation (Garzon et al. 2003).
1.2 A NIR Visitor Instrument
While the final suite of GTC instruments will sample all crucial optical and infrared
regimes, the first-generation of facility instruments do not include a NIR camera.
Professor Stephen Eikenberry designed the Canarias InfraRed Camera Experiment,
CIRCE, to fill the crucial gap between the telescope’s inception and the commissioning of
the second-generation NIR facility instrument EMIR and serves as principal investigator
(PI) for this project. While CIRCE is officially a “visitor” instrument, built solely with
funds from the University of Florida and the National Science Foundation, the instrument
is specialized for the GTC and is not intended to travel to other observatories. Further, as
opposed to traditional “visitor” instruments only used by the build team or with specific
permission from the instrument PI, the entire GTC community will be welcome to use
CIRCE. In return the CIRCE team receives “pay-back” time – guaranteed nights on the
telescope.
19
Besides functioning as a 1-2.5 µm broad-band imager, CIRCE will have the capacity
for narrow- band imaging, low- and moderate- resolution grism spectroscopy, and imaging
polarimetry. Other design features include fully cryogenic filter, slit, and grism wheels,
high-speed photometry modes, and broad-band imaging in J, H, and Ks filters (Edwards
et al. 2004, 2006). CIRCE’s 0.10′′ per pixel plate scale provides seeing limited images
even in the most excellent atmospheric conditions at the GTC site, estimated at ∼ 0.2′′
(Munoz-Tunon et al. 1997). Also, the planned 3.4′ × 3.4′ field is 25 times larger than
NIRC on the KECK Telescope (Matthews & Soifer 1994) and 3 times larger than NIRI on
the Gemini North Telescope (Hodapp et al. 2003).
In the first three chapters of this thesis, I describe the optical, opto- and cryo-
mechanical design of CIRCE. In Chapter 2, I show how CIRCE’s scientific goals and role
as a “workhorse” instrument influenced crucial features and specifications. I also present
the completed optical layout and our in-depth analyses of the optical design, concentrating
on characteristics such as enclosed energy and field-of-view. In Chapter 3, I discuss the
opto-mechanical design, starting with how I used the optical design to constrain all major
opto-mechanical components. I then review the mechanical design of brackets and the
optical bench. Finally in Chapter 4 , I detail my work on the cryo-mechanical design,
specifically the filter wheel or “pupil” box.
1.3 Science with CIRCE
While new instruments like CIRCE, EMIR, OSIRIS, and CanariCam explore the
“bleeding edge” of technology, it is important to remember that they are fundamentally
driven by science; observatories fund instruments that meet the needs of their users.
Occasionally, this requires commissioning a specialized instrument that will solve one or
two very specific scientific problems. More often scientists have an array of projects in
mind that drive the instrument. An observatory may organize an advisory committee to
discuss these broad goal and help define the instrument’s purpose.
20
CIRCE has such a committee, with membership spanning all of the GTC partners.
The CIRCE Advisory Committee members are: Drs. Stephen Eikenberry (principal
investigator) and Ata Sarajedini from the University of Florida, Dr Evenico Mediavilla
from the Instituto de Astrofisica de Canarias, Dr. Alberto Castro-Tirado from the
Instituto de Astrofisica de Andalucia, Mauricio Tapia from the Universidad Nacional
Autonoma de Mexico, and Dr. Peter Hammersley from the Gran Telecopio Canarias
Observatory.
Our team, with input from this advisory committee, designed CIRCE with a broad
range of scientific objectives in mind. Examples of projects outlined in the original NSF
grant proposal written by Dr. Stephen Eikenberry with contributions from the advisory
committee include: a campaign to observe microquasar GRS 1915+105 with high-speed
photometry and low-resolution spectroscopy, a polarimetric study of the relativistic jets of
microquasars, and JHK photometry of Galactic Globular clusters. These are only a few of
the myriad of projects one is capable of completing with CIRCE’s combination of modes
and features.
In this dissertation, I focus on another project that could hypothetically utilize
CIRCE’s capabilities; an emission-line search for massive star clusters around Soft Gamma
Repeaters (SGRs). While CIRCE’s FOV and narrow-band filter selection makes it
well-equipped for this survey, I did not use CIRCE to collect the observations presented
in this work. Given CIRCE’s anticipated completion in 2009, this was a strategic decision
made early in the project. By choosing to use the previous generation of NIR imagers to
obtain my observations, I was able to both participate in the early design work of CIRCE
and complete a viable science project within the timescale typical of a thesis. However, I
discuss how I plan to use CIRCE to continue the research presented in Chapters 5 and 6
in the future work portion of my conclusions (Chapter 8).
21
1.4 The Search for and Study of Massive Stars
Massive stars are defined as any star with an initial mass > 8 M¯ (Phillips 1999;
Heger et al. 2003a). During their life on the main sequence, they produce energy by fusing
hydrogen into helium via the carbon-nitrogen-oxygen (CNO) cycle. Unlike their less
massive counterparts, massive stars have convective cores; the bulk motion of gas particles
is responsible for heat transfer within their interiors. Massive stars are also supported, in
large part, by radiation pressure, making them less stable than stars like the sun (Phillips
1999).
While far outnumbered by low and intermediate mass stars, massive stars dominate
power output, CNO enrichment and ultraviolet (UV) radiation in the Milky Way (Morris
& Serabyn 1996). However, a large portion of the massive star population remains
undiscovered; their location in areas subject to high extinction and overcrowding, such as
the Galactic Plane, has barred optical surveys from detecting 95% of very massive star
clusters (Hanson 2003).
Despite lacking observations of significant portions of the Galactic massive star
population, intriguing discoveries of individual objects continue to surface, often posing
more questions than they answer. For instance, observations of the Pistol Star (Figer
et al. 1998), and LBV 1806-20 (Eikenberry et al. 2004b), with potential masses > 150
M¯, indicate that the maximum stellar mass may exceed the canonical theoretical limit
of 100 M¯ (Phillips 1999). Until we locate and study the majority of the massive star
population, the potential for discovering more “Very Massive Stars” that challenge our
current understanding of stellar physics is a reasonable possibility.
Furthermore, studies by Heger et al. (2003a) suggest that the deaths and resulting
progeny of massive stars may also defy our canonical picture of stellar evolution. Initially,
it was believed that stars with initial masses < 22-25 M¯ formed neutron stars while
stars with initial masses > 25 M¯ became black holes. New research indicates that this
picture may be too simplistic (Fryer 1999; Fryer & Kalogera 2001; Heger et al. 2003b).
22
During the late stages of their evolution, massive stars often have strong stellar winds that
result in significant mass loss. Thus, objects originally destined to end their lives as black
holes may lose enough mass during pre-supernovae Wolf-Rayet (WR) or Luminous Blue
Variable (LBV) phases to instead evolve into neutron stars (Heger et al. 2003a,b).
With these interesting problems in mind, we proposed a survey to search for massive
stars. In particular, we chose a “targeted” approach, probing for massive stars around
interesting objects potentially associated with young clusters; in this case Soft Gamma
Repeaters (SGRs). SGRs are magnetars – a rare type of neutron star (Duncan &
Thompson 1992). Besides the potential to increase the size of the known massive star
population, we hoped our investigation of SGR environments might shed light on the
astrophysical problem of the origins of magnetars and the fates of windy massive stars.
In the following sections, I review magnetars, previous observations of their
environment and the mechanics of our search for massive stars in the following sections.
1.5 Observations of Magnetars
On March 5, 1979, a giant gamma-ray burst, peaking at 1045 ergs s−1, flooded every
gamma-ray detector in orbit around Earth. The hard spike was followed by a tail of
soft-gamma ray bursts that varied with an 8s period (Mazets et al. 1979; Golenetskii
et al. 1984) and lasted three minutes. In the following years, the same source, located
in the Large Magellanic Cloud, emitted a number of lower energy soft-gamma bursts
lasting from 50 ms to 3.5s (Golenetskii et al. 1984). The repetitive nature, absence
of high-energy gamma-rays, and duration of these bursts was similar to bursts seen
subsequently in two other anomalous Galactic sources (Laros et al. 1987). Notably
different from classic gamma-ray bursts (GRBs) which do not repeat, the objects earned
their own classification: Soft Gamma Repeaters (SGRs). The LMC source was designated
SGR 0526-66; the other two are SGR 1806-20 and SGR 1900+14.
While a variety of hypothesis to explain these objects abounded in the following
two decades, the most successful theory, developed by Duncan & Thompson (1992)
23
is the magnetar model. In this scenario a newborn neutron star rotating very quickly
creates a dynamo that strengthens the stars magnetic field. As the star cools, the dynamo
action stops; however, the magnetic field, which has reached 1014 - 1016 G, is locked
into the star’s dense interior. Duncan & Thompson (1992) believed that instabilities in
this strong magnetic field created both the large and small soft gamma-ray flares. They
also predicted that the dissipation of rotational energy by magnetic waves would slow
a magnetar in the first few seconds of its birth, resulting in a rotation period of ∼ 10s.
Further observations of SGRs led to the discovery of persistent X-ray emission (Murakami
et al. 1994; Rothschild et al. 1994; Vasisht et al. 1994), explained by Thompson & Duncan
(1996) as decay of the magnetic field.
As relayed in Woods & Thompson (2006), significant confirmation of the magnetar
model came in 1998 with the discovery of a 7.5 s period and 0.0026 s year−1 spin-down in
the X-ray emission of SGR 1806-20. The spin-down was attributed to magnetic braking
of a neutron star with a 1015 G magnetic field (Kouveliotou et al. 1998a), in the range of
that predicted by the magnetar model (Duncan & Thompson 1992). Then, on the heels
of this announcement, Hurley et al. (1999) reported observations of a giant flare from
SGR 1900+14 , similar in morphology to the earlier SGR 0526-66 event (see Figure 1-2).
This effectively settled any doubt that SGR 1900+14 and SGR 0526-66 were indeed the
same class of object. Soon after, SGR 1627-41 was added to the roster of SGRs after
Kouveliotou et al. (1998b) and Woods et al. (1999) reported 100 soft gamma-ray bursts
emitted from the source. This brought the total number of SGRs to four, where it has
remained since.
There is significant evidence that another flavor of neutron star, known as Anomalous
X-ray Pulsars (AXPs), are also magnetars. First discovered by Fahlman & Gregory
(1981), AXPs exhibit fast spin-down rates, 5-7 s periods, and 1035 - 1037 erg s−1 X-ray
emission. Isolated objects, they were originally termed “anomalous” due to the absence of
a reasonable mechanism to explain their high X-ray luminosity. However, while accounting
24
for similar X-ray luminosities in SGRs, Thompson & Duncan (1996) proposed magnetic
decay as the source of AXPs’ energy and included these objects in their consensus of
magnetars. This interpretation was strengthened by Gavriil et al. (2002) and Kaspi &
Gavriil (2003) with the discovery of SGR-like bursts from two AXPs, 1E 1048.1-5937 and
1E 2259+586 (see Figure 1-3). Currently, astronomers recognize 8 AXPs and candidate
AXPs.
In summary, SGRs exhibit: [1] “giant” (∼ 1044 ergs/s) gamma-ray bursts characterized
by a hard spike and soft-gamma ray tail with variations on the order of 10s (Figure 1-2),
[2] repetitive, bright (∼ 1041 erg/s) soft-gamma ray bursts lasting ∼ 100 ms (Figure
1-3) and [3] persistent X-ray emission with a typical luminosity ∼ 1035 - 1037 ergs/s.
Less common and not well understood are “intermediate” SGR bursts, which display
characteristics of both the repetitive soft gamma-ray and the giant bursts (Woods &
Thompson 2006, and references therein). AXPs exhibit [1] SGR-like soft gamma-ray
bursts and [2] persistent X-ray emission similar to SGR X-ray emission. To date, no
“giant” AXP flares have been observed.
Magnetars are set apart from other neutron stars on the spin period vs. period
derivative (P-P ) diagram. Shown in Figure 1-4 (Woods & Thompson 2006), a P-P
diagram groups neutron stars according to their periods (P), change in periods over time
(P ), and magnetic fields (B). Classical neutron stars are located in the central region
of the diagram. Magnetars, which spin down more quickly and have longer periods and
stronger magnetic fields than classical neutron stars, populate the upper left corner.
1.6 Environments of Soft Gamma Repeaters (SGRs)
Magnetars offer a unique opportunity to investigate the exotic endpoints of stellar
evolution. With only a handful of confirmed members plus a few additional candidates
known, the potential to study the entire population of magnetars, establish formation
scenarios, and characterize their progenitors is a tantalizing and realistic objective.
25
To this end, several authors have performed studies focused on magnetars and their
environments. Interestingly, of the four SGRs known, three are potentially associated with
massive star clusters. SGR 1806-20 is embedded in a cluster of massive stars, Cl 1806-20
(Fuchs et al. 1999; Corbel & Eikenberry 2004; McClure-Griffiths & Gaensler 2005; Bibby
et al. 2008) that contains both a Wolf-Rayet star and LBV 1806- 20, potentially the most
massive stellar object discovered to date (Eikenberry et al. 2004a). Vrba et al. (2000)
reported the discovery of a potential massive stellar cluster with at least 13 members a few
arcseconds from SGR 1900+14; recent observations by Wachter et al. (2008) suggest that
the cluster and magnetar are at the same distance. Finally, Klose et al. (2004) reported
the discovery of a young cluster of stars in the vicinity of SGR 0526-66.
This observational evidence, combined with recent theoretical models developed by
Heger et al. (2003a) raises an interesting question: if magnetars are uniformly associated
with clusters of massive stars, might we conclude that their progenitors were also very
massive? Magnetars are compact objects formed from stars that have already progressed
through most of their evolutionary stages. Since stellar evolution dictates that in order
to have already reached the endpoint of their stellar lifespan, these progenitors must be
more massive than the most massive stars currently in their natal clusters, this suggests
scenarios in which exceptionally massive stars explode to form not only neutron stars, but
particularly, highly-magnetized neutron stars (Eikenberry et al. 2004a; Figer et al. 2005;
Muno et al. 2006).
The existence of massive stars and neutron stars in the same cluster could also be
evidence of multi-epoch star formation (Eikenberry et al. 2004a). This theory proposes
that shocks from the explosive death of one star might induce a wave of star formation,
yielding a new generation of stars. Observational evidence for this mode of triggered star
formation reaches back for decades and encompasses several well-known OB associations
including Sco OB2 (Preibisch & Zinnecker 2001) and CepOB3 (Assousa et al. 1977; Pozzo
et al. 2003).
26
While these possibilities are intriguing, one must first address a number of factors
before developing either idea into a fully realized and corroborated scenario. SGR 1806-20
is currently the only SGR clearly linked by distance measurements to a surrounding
cluster of massive stars, Cl 1806-20 (Corbel & Eikenberry 2004; McClure-Griffiths &
Gaensler 2005; Bibby et al. 2008). Kaplan et al. (2002) raised concern over whether
SGR 1900+14 was a member or former member of its nearby cluster and the cluster
membership of SGR 0526-66 has yet to be established (Klose et al. 2004). I could find no
effort in the literature to find a cluster associated with SGR 1627-41.
To investigate the potential association of SGRs with massive clusters I utilize
near-infrared narrow-band imaging to search for emission lines indicative of stellar winds
in massive stars (Figer et al. 1997; Hanson et al. 1996). Specifically, I use a narrow-band
filter centered on the Brγ 2.16µm line, or a He I filter centered on the 2.058 µm line.
In extremely massive stars, these emission lines are very pronounced. For instance, the
equivalent width of the Brγ line in one previously identified WR in Cl 1806-20, Star 22 in
LaVine et al. (2003) (also known as Star 2 in Figer et al. 2005), exceeds 40A; the He I line
in Star B is > 100A.
These prominent emission lines may be useful in crowded fields, identifying massive
stars even when they are scattered amongst many field stars. As opposed to other
emission lines characteristically used to identify massive stars, most notably Hα in the
optical, the Brγ and HeI lines are located in the near-IR and therefore detectable even
in regions highly affected by interstellar extinction (such as Cl 1806-20). This study may
also yield detections of OBI cluster stars with Brγ absorption. While less marked than
reported Brγ emission lines in WRs, Brγ absorption may reach EW = 5A in some OBI
stars (Hanson et al. 1996).
We present the results of our narrow-band imaging search for massive stars around
SGRs in Chapter 5 and 6. In Chapter 5 we detail the observations, analysis, and
photometric techniques used in this study, particularly as they apply to Cl 1806-20.
27
In Chapter 6, we will present results on our narrow-band imaging of the other three SGR
environments.
1.7 The Locations of Luminous Blue Variables (LBVs)
While researching Westerlund 1 and Cl 1806-20, two Super Star Clusters associated
with magnetars, I noted that both clusters also hosted a Luminous Blue Variable (LBVs),
a rare type of evolved massive star. While examining finder charts for both clusters, I
serendipitously noticed that both LBVs are on the outskirts of their respective clusters.
Upon further investigation, I found that the LBVs in the Quintuplet, another SSC, are
also not centrally located. In fact, four of five LBVs in young massive clusters appear to
be located on the periphery of their host clusters. The exception to this observation is Eta
Carina in Trumpler 16.
LBVs are also thought to be at the extreme of the mass spectrum; in fact Eta
Carinae and the Pistol Star and LBV 1806-20 vie for the title of most massive star in the
Galaxy each estimated at > 120 M¯ (Figer et al. 1998; Eikenberry et al. 2004a). Clearly,
given the extreme masses of LBVs and the dictates of mass segregation one would suppose
that LBVs would be centrally located (Spitzer 1987; Binney & Tremaine 1987). Theory
suggests that LBVs should not be on the outskirts of their cluster, which is where I found
them upon my initial visual inspection.
In Chapter 7, we evaluate the status of LBVs as “peripheral” cluster members. We
calculate the probability of finding at least one star in every cluster at or outside the
radius of the resident LBV and determine the robustness of our result using a Monte
Carlo simulation. Finally, we compare our finding to theoretical models that predict the
locations of massive stars in clusters and contrast these with scenarios that reproduce the
observed positions.
28
Figure 1-1. The mirrors and foci of the GTC. M1 is the 36 segment primary mirror, M2 isthe lightweight beryllium secondary mirror, and M3 is a movable tertiarymirror. When the tertiary is not in the optical path, light will travel to theCassegrain Focus (CF). When it is in the path, light can be directed to anyone of the four Bent or Folded Cassegrain Foci (FCF) or one of the twoNasmyth Foci (NF)
29
Figure 1-2. The 1998 August 27 observations of the giant gamma-ray burst from SGR1900+14 as reported by Hurley et al. (1999, see Figure 1). Note the brightspike followed by a long tail. The period of the oscillations in the tail is ∼ 5s
30
Figure 1-3. This figure, from Gavriil et al. (2002, see Figure 1) shows a typical soft-gammaray burst from a magnetar. This burst was from 1E 1048-5937, an AXP;however, SGR bursts have almost identical morphologies.
31
Figure 1-4. This figure from Woods & Thompson (2006) shows the class spin period vs.
period derivative (P P ) diagram for neutron stars. Period derivative is definedas the change in the period over time. The diagram also has lines denotingmagnetic field strength. Classical neutron stars are located in the centralregion of the diagram. Millisecond pulsars are located on the bottom left.Magnetars are located at the upper right. They spin down more quickly andhave longer periods than classical neutron stars. They have magnetic fields ∼1014 - 1016 G.
32
CHAPTER 2OPTICAL DESIGN AND ANALYSIS OF CIRCE
As with all astronomical instruments, the design of the Canarias InfraRed Camera
Experiment (CIRCE) was heavily influenced by the science interests of its intended users
– in this case the consortium of GTC astronomers. As the only near-infrared (NIR)
instrument planned for the early phases of the GTC’s operation, CIRCE was envisioned as
a “workhorse” camera, capable of meeting the NIR needs of the entire community.
To satisfy this criterion, CIRCE was conceptualized with two basic modes, imaging
and spectroscopy. A third mode, polarimetry, was added to ensure that CIRCE would
remain scientifically viable after EMIR, the facility NIR imager and spectrograph, is
commissioned on the GTC. CIRCE’s optical design was strongly influenced by these
early decisions as well as a number of specifications intended to optimize the instrument’s
performance. In this chapter, I review the various stages of CIRCE’s optical designs. I
discuss the initial specifications and their impact on the layout, changes implemented to
meet GTC requirements, and our analysis of the system.
2.1 Preliminary Optical Designs
The earliest CIRCE optical layout, created before I joined the CIRCE team, was an
all-refractive design. All-refractive systems use only lenses as powered optics – elements
that change how the beam of light from the telescope converges or diverges. The initial
CIRCE design, shown in Figure 2-1, consisted of 8 lenses made of a variety of materials
including Barium Fluoride and Zinc Selenide.
Often, designers employ all-refractive systems (e.g. WIRC (Wilson et al. 2003)
and Flamingos-1 (Elston et al. 2003)) because they are the simplest option, allowing
light to travel from lens to lens along a single axis. However, in the case of CIRCE the
all-refractive design, which called for expensive anti-reflective coatings and high-index
glass, was costly, overly-complex, and less stable in cryogenic environments. Further, the
33
expected throughput, the amount of light propagated through the optical system, was only
average since light from the telescope passed through 8 lenses before striking the detector.
When I joined the CIRCE team, the CIRCE PI was considering another option; an
all-reflective system, which utilizes only mirrors as powered optics. Besides promising
better throughput for less money (see §2.5), a reflective design also solved a major
problem associated with cryogenic environments. As with most NIR instruments, CIRCE’s
components will be enclosed in a vacuum vessel, or dewar, cooled to 77 K. While some
lenses and coatings are designed for these extreme conditions, damage to lenses is still
possible. As temperatures drop inside the dewar, lenses, composed of a variety of
materials, contract differently than their aluminum mounts. If a lens is accidentally
compressed by a misaligned or poorly designed mount, it will crack or shatter.
Mirrors do not face this problem. Diamond-turned from 6061-T6 aluminum (see
3.1), the same material used to make all other opto-mechanical components (described
in Chapter 3), the optics have the same coefficient of thermal expansion as the bench
and optical mounts. Thus, the entire system undergoes homologous contraction. Besides
preventing damage to expensive optical components, homologous contraction also means
that a system aligned while warm will remain aligned when cooled; a significant advantage
during integration. In light of these benefits, the CIRCE team decided to pursue mirrors
instead of lenses for CIRCE’s optical design.
This decision was bolstered by the successful implementation of all-reflective optics
by the InfraRed Astrophysics Group (IAG) in the Department of Astronomy at the
University of Florida (UF). The IAG successfully installed and aligned reflective optics in
two mid-infrared instruments, T-ReCS and CanariCam (Telesco et al. 2003, 1998) and a
NIR Integral Field Unit (IFU), FISICA (Eikenberry et al. 2004a). While diamond-turned
aspheres have been historically difficult to align and test, advances in manufacturing
resulted in a “bolt and go” technique that nullified alignment issues. Given the experience
of the engineers and instrument scientists of the IAG, who acted in an advisory role
34
throughout the design and manufacture of CIRCE, we felt that using mirrors was not only
feasible, but the best available option.
With this in mind, the CIRCE PI, Professor Stephen Eikenberry, contracted Optical
Research Associates (ORA) in Pasadena, California to perform a feasibility study. The
PI developed a number of design specifications that would accommodate the desired
observing modes (imaging, spectroscopy, and polarimetry), minimize optical aberrations,
and meet design and envelope constraints imposed by the GTC (§2.3).
Several of these specifications influenced the optical design at the most elementary
level. First, we required that the system contain an accessible exit pupil, where we
could place a “cold stop”. A cold stop is a mask placed into the optical path that
stops scattered light from propagating to the detector focal plane. We also specified a
large collimated airspace to accommodate several filter and grism wheels. This avoided
positioning filters near the detector focal plane, where dust and filter defects would
be projected onto the detector. The easiest way to meet this requirement was with a
collimator/imager system; a common configuration for IR instruments. I detail this system
further in §2.4.
We limited the size of the beam at the exit pupil to 55mm in diameter to minimize
filter and grism costs, which increase dramatically with physical size. We set the
pixel scale at 0.1 ′′ per pixel to sample the optimal 0.2 arcsec seeing at the GTC site
(Munoz-Tunon et al. 1997) and required a 3.3′ × 3.3′ field-of-view (FOV). At the time
of the design study, this FOV was considerably larger than that offered by similar NIR
instruments available on 8-10 meter class telescopes. I discuss the CIRCE FOV further in
§2.5
After reviewing several proposed optical layouts presented by ORA optical designer
Michael Rodgers, the final result of the feasibility study was the choice of a two mirror
collimator with a four-mirror camera. This design had the best performance of the
35
systems presented (discussed in detail in §2.5) and met the abovementioned design
specifications.
Once we chose the basic layout, ORA completed an in-depth study of the system.
To avoid spherical aberration, the initial design used a variety of higher-order aspheres
(see Equation 2–1). Since these are difficult to manufacture, the CIRCE PI requested
that ORA modify the imager to minimize the number of aspheres. ORA incorporated
this consideration into the next integration. I present the completed preliminary design, a
2-dimensional layout of CIRCE in Figure 2-2.
2.2 Introduction to Optical Design Software and Terminology
Before the advent of modern computing, tracing a single ray through an optical
system could take many days of careful calculation. Now, optical designers can choose
from a variety of software to complete hundreds of ray traces, analysis, and optimization
in a few hours (Smith 2005). Throughout this dissertation, I will discuss work completed
with two such programs, CODE V and ZEMAX.
As the manufacturers of CODE V (Harris 1991), ORA chose this program to model
their preliminary designs of CIRCE. Since ZEMAX was available at UF, I translated the
CIRCE optical prescription from CODE V to ZEMAX. This required learning optical
design principles and the intricacies of both programs. While I will not detail the latter,
I will discuss the former, reviewing coordinate systems and terminology referenced
throughout Chapters 2 and §3.
In standard practice, the coordinate system used in optical design is the right-handed
Cartesian system shown in Figure 2-3 (Schroeder 2000). The Y-axis is vertical and
increases from bottom to top. The X-axis runs into and out of the page and increases into
the paper. The horizontal Z-axis increases from left to right. In layouts where all optics
are coaxial (aligned on a single axis) the Z-axis aligns with the optical axis, the imaginary
line along which light moves or propagates through the system. This is not the case with
CIRCE, where many optics are off-axis.
36
Examining Figure 2-2 in terms of this coordinate system, note that we are looking at
the optics as though they were suspended directly above us. Further we are only observing
a two-dimensional (2-D) cross section taken at X = 0; practically speaking we see each
optic’s length and width, but not its height. This point-of-view will become particularly
important in Chapter 3.
Next I move on to a few basic optical design concepts. Both CODE V and ZEMAX
have spreadsheet-like interfaces that allow the user to enter information about each optical
element. Figure 2-4 is an example of the CIRCE optical prescription displayed in the
ZEMAX interface. Of particular import are the following user input variables: surface
type, glass, thickness, decenter, tilt, and aperture decenter.
Surface type is a mathematical description of an optic’s face. As discussed in §2.1,
most of CIRCE’s optics are aspheres, used specifically to eliminate spherical aberration.
Aspheres are described by the following equation:
Z(R) =UR2
√1− (1 + K)C2R2
+ A4Y4 + A6Y
6 + ...AnYn (2–1)
where U is the radius of curvature, K is the conic constant, R is the height from the
optical axis, Z(R) is the sag, or shape of the surface at height R, and An are the aspheric
coefficients. Five of CIRCE’s six powered optics are conic aspheres with only a conic
constant, K, and no higher order terms. The fourth mirror in the imager is a sixth order
even asphere, with two aspheric coefficients, A4 and A6.
A coordinate break is another important surface type. It is a false or dummy surface
that allows for the off-axis placement of optical components. A coordinate break defines a
new coordinate system in terms of the coordinate system currently in use. Decenters shift
the new system in X and Y, while tilts rotate the new system in X, Y, and Z. Tilts around
X are defined as α, Y as β, and Z as γ.
Thickness is the distance along the local Z-axis from one surface to the next.
Practically, this may have different physical meanings depending on the type of optic.
37
In the case of filters, lenses, and entrance windows, the thickness between the front
and back face of the optic is the width of the element. In the case of mirrors, with only
one face, the thickness is the distance between one mirror and the next. The “glass”
designation therefore denotes the physical meaning of the thickness associated with a
surface. Reflective optics are listed as mirrors while the composition of refractive optics
is explicitly stated. The optical properties of thousands of materials are stored within the
optical design software.
All of CIRCE’s powered mirrors are segments of larger aspheric optics called “parent”
mirrors. For each optic in a complex system, an optical designer chooses the radius of
curvature (RC) necessary to correct aberrations and direct light toward the next mirror
in the optical path. The RC sets the size of the optic; a larger radius of curvature results
in a large optic. Using a mirror approximately the size of the beam, as shown in Figure
2-5a, may yield a radius of curvature too small to correctly reflect the light to the next
optic. Figure 2-5b demonstrates that although a large RC, and therefore larger mirror,
is necessary, only a portion of this “parent” mirror is needed to reflect the beam of light
directed at it by previous optics. The remainder of the mirror not only adds cost to the
project, but might block light intended for another optic. Thus, we use the smallest
segment of the parent mirror that still accommodates the entire beam of light. Finding
this optimum size is discussed in Chapter 3.
The two collimating mirrors in CIRCE are off-axis sections of coaxial parent mirrors.
The four imaging mirrors are smaller segments of non-coaxial parent mirrors. Figure 2-6
shows the parent mirrors used in the design of CIRCE. When using sections of parent
mirrors, we define the aperture decenter as the distance between the vertex of the parent
mirror and the mechanical center of the mirror segment. Aperture decenters should not be
confused with coordinate break decenters.
As a complex off-axis optical system, the CIRCE layout has many coordinate breaks,
aperture decenters, glasses, and thicknesses. A complete prescription for the preliminary
38
CIRCE optical system, which defines all optical elements and coordinate breaks, is listed
in Table A-1 and A-2.
2.3 Envelope Re-design and GTC Integration
As CIRCE moved forward from the preliminary design phase, I became more involved
with the project and assisted with several stages of the optical design, beginning with an
analysis of the optical layout. One important consideration was the GTC envelope size.
CIRCE was designed for the Bent Cassegrain port of the telescope, which has specific
weight and size constraints. While CIRCE fit into the 1.0-m × 1.5-m cylindrical volume
planned for the Bent Cassegrain port, further investigation of the attachment points,
shown in Figure 2-7, uncovered a problem. The telescope’s optical axis is centered on the
1.0-meter cylinder, allowing for only a 0.5-m clearance to the farthest optic on the Y-axis.
If we aligned CIRCE’s entrance window (located at the upper left of the layout in Figure
2-2) to the telescope’s beam, CIRCE’s farthest optic would be 1-m from the optical axis.
Thus the size of the cylinder necessary to contain the optics would be 2.0-m × 1.5-m,
double the diameter specified by the GTC (Figure 2-7).
I contacted ORA with this information, while simultaneously beginning to alter
the CIRCE design. My initial idea was to add fold mirrors in critical places along the
optical path. Fold mirrors are unpowered optics and therefore only change the direction
of the light as it travels through the system. They are especially useful in situations that
require fitting an existing optical design into a smaller space without radically altering the
position of the powered mirrors.
In this case, I needed to compress the CIRCE design in the vertical (Y) direction.
In Figure 2-8, I show my first re-design, consisting of 10 mirrors: two collimators, four
imagers, and four fold mirrors. The first set of fold mirrors translated a large thickness
between the entrance window and the collimating optics into an offset in the positive Y
direction. I added a second set of fold mirrors aft of the collimator that reflected the light
39
off of the X = 0 plane into the negative X direction. In this plan, two optical benches
supported the optics instead of one.
While this design met the envelope specifications detailed by the GTC it faced several
other problems. Foremost, it lacked space for a large filter box. With a variety of filters
and grisms planned for the instrument, we anticipated a multi-wheel filter box aft of the
collimator. Concerns that the additional fold mirrors would greatly limit our access to the
collimated airspace necessary for the filter box dominated discussion of this new layout.
Despite these concerns, I sent our 10-mirror CIRCE concept to ORA. After several
iterations, ORA delivered a final design which I present in Figure 2-9 . While including
the fold mirrors fore of the collimator, as in our design, this new layout utilized a slightly
altered four-mirror imager to meet the envelope specifications. Specifically, the modified α
tilt on the first imager directed the light upwards toward positive Y values. This greatly
compressed the design while leaving the collimated airspace necessary for the filter box
virtually unchanged. Deleting two fold mirrors from our design also decreased the cost and
removed the second optical bench.
Once the final design was complete I worked with ORA to integrate the CIRCE
optical design with that of the GTC. Members of the CanariCam and Elmer teams
provided me with the ZEMAX spreadsheet for the GTC. Since the primary and secondary
mirror of the GTC are polygons, both teams used complex user-defined surfaces contained
in file types specific to ZEMAX to model the GTC optics. To share these designs
with ORA (who used CODE V) I simplified the schematics of the GTC optics while
maintaining the basic optical properties. This new GTC design was sent to ORA and all
future work and analysis used this layout.
I then mated the CIRCE optical design to the GTC optical design using the
telescope’s focal plane. I modeled the telescope focal plane in the CIRCE design as a
“dummy” surface, placed directly after the entrance window. While this dummy surface
had no optical properties, it marked the position of future mechanical mechanisms. After
40
taking account slight shifts in the location of the focal plane due to the entrance window, I
matched the final focal plane of the GTC to the focal plane in the CIRCE design.
2.4 Final Optical Design
I now review the final optical design presented in Figure 2-9. An F/17 beam from
the telescopes enters the CIRCE entrance window [1] before passing to the GTC focal
plane [2]. As discussed in §2.3 this focal plane marks the location of the first of CIRCE’s
cryo-mechanisms, a device to move slits and field masks in and out of the beam.
After passing through the focal plane mechanism, light will strike two unpowered fold
mirrors before passing to two collimating conic mirrors; the first concave and the second
convex. As with the preliminary design, both are off-axis segments of coaxial parent
mirrors. These optics produce an exit pupil, the image of the aperture stop, which in
this case is the GTC secondary mirror (Schroeder 2000). The ∼ F/5 beam is ∼ 55mm in
diameter at the exit pupil. As previously mentioned, to stop stray light from propagating
into the camera, we will place a mechanical cold stop at this position with this dimension.
Aft of the collimator is ∼ 300mm of collimated airspace designed to accommodate the
second cryo-mechanism, a large filter wheel box housing a cold stop [3], filters, grisms, and
optics for polarimetry [4].
Next is the four-mirror imager. The first imaging mirror is a concave conic mirror,
the second and third imaging mirrors are convex conic mirrors, and the fourth a concave
6th-order asphere. Neither the parent or segments of the imaging mirrors are co-axial.
While this will make manufacturing more difficult, the extra tilt and decenter parameters
increased ORA’s ability to minimize aberrations when optimizing the design.
The camera focuses light onto the detector focal plane [5], where we will position a
mercury cadmium telluride (HdCdTe) HAWAII 2048 × 2048 infrared detector.
2.5 Optical Analysis of Imaging Mode
With the final design complete, I began an in-depth optical analysis of the system. I
used several key utilities in ZEMAX to predict how well CIRCE would perform. First, I
41
checked the CIRCE field-of-view (FOV). Using ZEMAX, I virtually launched rays into the
telescope pupil that corresponded to a 3.4′ × 3.4′ square on sky, the expected FOV of the
system. I then set the aperture size at the final focal plane equal to 36.86mm × 36.86mm,
the size of the the HAWAII 2K detector. Using the FOOTPRINT utility in ZEMAX,
which shows where each propagated ray will hit any surface in the design, I examined the
focal plane. In Figure 2-10, I show the results. The marks on the diagram represent rays
launched from 8 locations on the pupil plane that mark the boundaries of a 3.4′ × 3.4′
square on the “sky”. This defines the CIRCE FOV.
I then analyzed the system’s encircled energy. Encircled energy (EE) is the
percentage of total energy enclosed as a function of distance from the center of the
PSF. With this analysis we can predict how much light CIRCE will collect within a
given number of pixels at the detector focal plane. Since the HAWAII 2K detector has
18µm pixels, I was interested in the encircled energy at a radius of 18µm, or two pixels in
diameter, which guarantees Nyquist sampling (Schroeder 2000).
I present the results of my encircled energy analysis for the J- and K- band in
Figures 2-11 and Figure 2-12, respectively. Figure 2-11 shows that at the center of the
field (0, 0) the encircled energy is > 90% in two pixels. At ∼ 1.5′ from the center (0.250,
0.250) the EE within 2 pixels is 80% and in the corner (-0.028, -0.028) decreases to 75%.
This slight drop-off in EE at the corners was expected; we requested ORA to optimize
the image quality over a 205′′ field diameter, sacrificing a uniform field for excellent
throughput (80% EE in two pixels) over > two-thirds of the detector. We note that the
EE over the entire field is significantly better than the original lens design for CIRCE,
which had only 60% EE over two pixels.
Figure 2-12 shows the EE diagram for the K-band. Note that the percentage of
energy enclosed at each position is slightly worse than in the J-band. This was expected
since diffraction, and therefore spotsize increases with wavelength, however, we still have
42
> 70% EE at two pixels, even in the corner, the worst part of the field. In the center (0,0)
we reach > 90% EE within two pixels.
I then performed a spot size analysis shown in Figure 2-13. This shows where rays
from a given field position will fall on the detector focal plane. The theoretical Airy Disk
is also plotted for comparison. The spot size is a measure of how close to ideal the optical
system is to the diffraction limit; generally the tighter, more uniform and circular the
spot, the better the design. Spot diagrams also provide insight into the major sources
of aberration in an optical system. For example, coma appears as a “comet”, spherical
aberration as a circle with a tight core and underdense outer region, and astigmatism as
an exaggerated oval. In systems with chromatic aberration, which is the most obvious
aberration, different wavelengths appear as spots with varying sizes and shapes.
Often, more complex systems like CIRCE have a combination of these effects, along
with other higher-order aberrations, which make it difficult to determine what is causing
the shape of the spot. However, when examining Figure 2-13 I note no obvious signs of
chromatic or spherical aberration. While there may be some coma, the spot size is very
small, mostly within the Airy radius.
The analysis of the CIRCE spot diagram confirms the validity of our design
specifications. Since CIRCE uses mirrors instead of lenses and aspheres instead of spheres
we eliminated spherical and chromatic aberration by design. Further, specifying excellent
image quality (measured in terms of encircled energy) across the bulk of the field ensured
a tight spot size; rays must land close together to ensure 80% energy within two or three
pixels.
Finally, we examined the distortion in the system. If one imagines a grid of lines,
distortion is a measure of how the lines diverge or warp as distances increases from
the optical axis. In a system with no distortion, the lines remain straight. If distortion
is extreme the lines curve and the field takes on a barrel or pincushion shape. Our
specifications called for less than a 10 pixel distortion in a 205′ radius from the center
43
of the image. Our analysis in ZEMAX showed that CIRCE exceeded this specification,
experiencing 4 pixels of distortion in the 205′ radius.
2.6 Spectroscopy and Polarimetry
To this point, we have reviewed the general optical properties of the optical design
and analyzed CIRCE’s performance as an imager. CIRCE has two additional modes,
spectroscopy and polarimetry.
CIRCE spectroscopy includes two grisms, both of which are derived from designs for
the Flamingos-2 instrument for Gemini. To save time and money, the CIRCE grisms share
the grating ruling masters custom-built for Flamingos-2. The first grism will cover the
1.25 - 2.4 µm bandpass at a resolution of R = 410 at 1.25 µm and R = 725 at 2.20 µm
with a three pixel slit. The second grism will cover a single band at a resolution of R ∼1500 in its 3rd order (K-band), 4th order (H-band), or 5th order (J-band). The CIRCE
optics will maintain seeing-limited image quality with the grism in the optical path over
the entire bandpass.
CIRCE will also have a Wollaston prism polarimetric mode. In this mode, a
Wollaston prism will be placed in the beam after a rotating half-wave plate (HWP)
located at the focal plane. This prism will deviate the ordinary and extraordinary
polarization beams at the pupil, resulting in a shift of the “o” and “e” images on the
detector. Coupled with a half- size field mask also at the telescope focus, this will provide
spatially-separated images, allowing polarimetric measurements. A rotating HWP,
positioned at angles of 0-degrees and 45-degrees allows the measurement of the “Q”
polarization component. Rotating the plate to 22.5-degrees and 67.5-degrees allows the
measurement of the “U” component, and thus the complete determination of linear
polarization in the source. Finally, by placing a grism in the beam after the Wollaston
prism, CIRCE will also be capable of spectropolarimetry. Another member of the CIRCE
team, Miguel Charcos, is leading the polarimetric design.
44
2.7 Tolerancing Analysis
Up to this point, the analyses that I have reviewed in §2.5 assume perfect optical
elements. While the precision and accuracy used to manufacture diamond-turned optics
is very high, small defects invariably occur during machining and alignment. These errors
may adversely affect the final image quality and render our carefully completed analyses
inaccurate. Fortunately, we can specify a maximum acceptable error, or tolerance, on
every dimension used to produce the mirror. These include the X, Y, and Z location of the
mechanical center and the radius of curvature. If an optic returns from the manufacturer
with errors larger than those declared on the drawing, we can reject it. The manufacture
is then obliged to re-make the piece.
Deciding what range of errors are acceptable is a delicate process. It is critical to find
the right balance between quality and cost. If a manufacturer agrees to produce a piece
with very small or “tight” tolerances, they assume a large risk. Besides guaranteeing a
near-perfect optic, they are also responsible for testing the optic to prove they have met
the tolerance. Obviously, the tighter the constraints, the more difficult the manufacture
process, the greater the expense. However, if tolerances are too “loose” and the optics
poorly manufactured, the instrument could be unusable, negating any benefit gains
from reducing costs. Also, a limit in machining accuracy does exist; we are ultimately
constrained by the tools and technologies available.
To add realism to our system and determine the tolerances for our CIRCE optics for
the opto-mechanical design, I completed a tolerancing analysis. Using macros written by
ORA, based on an analytical technique detailed in Koch (1978), I ran an iterative process
to identify specific parameters that impacted image quality. By deciding which variables
in our optical design had large effects on two measures of system quality, distortion and
encircled energy, I determined which parameters required tight tolerances. This would
ultimately help determine CIRCE’s cost and manufacturing difficulty.
45
The basic parameters that I adjusted were tilt, decenter, curvature, aspheric
irregularities, and window and fold mirror materials: 45 variables in all. Within the
macro, I could specify tolerances for each parameter and a list of field positions, similar
to those used for the initial EE calculation. CODE V then performed the analytical
calculation and output two tables. The first measured the diameter of a circle necessary
to enclose 80% of the energy. This was similar to ZEMAX’s encircled energy analysis
described above; however, in this case, the percentage is held constant and the diameter
in microns is reported. The smaller the diameter, the smaller the spot size. The second
analysis measured the distortion (in pixels) at each field position.
Each table had five values of the 80% EE or distortion for each field position. The
first was the nominal value, assuming no errors in the system. The next four values were
the expectation values for four confidence levels, 50%, 84.1%, 97.7%, and 99.9%.
For example, for the initial run, I stipulated a tolerance on the α and β tilt of every
mirror in the system equal to 0.001 degree a typical tolerance for optical alignment. The
results are shown in Table 2-1. Originally, we found that at the center of the field (0,0)
the nominal value of the 80% EE was 25µm in diameter, about 1.5 pixels. However,
assuming that the tilt could be as much as 1 µm larger or smaller than the nominal value,
the 50% value was 35µm; 50% of the time we could guarantee an 80% EE in 2 pixels or
less. The remaining 50% of the time the value would be larger, most likely in situations
where multiple tilts were off by the maximum amount. On the other hand, 84.1% of the
time, we guaranteed the 80% EE would be 36.30µm or less, 97.7% of the time, 41.9µm
or less, and 99.9% of the time 47.50µm or less. Recalling that the center of the field
often has considerably better image quality, these values are rather large in comparison
to the nominal value. We determined that this tolerance was not “tight” enough; if we
aligned the optics to this level of error, our image quality would most likely be poorer
than originally specified.
46
After completing multiple runs of the program like the one detailed above, I identified
the parameters that most affected the performance of the system. I also set benchmarks
for acceptable image quality threshold. At all field positions, I sought an 80% EE of 32µm
and a distortion of < 4.4 pixels at the 84.1% confidence level. After ∼ 10 runs, I found a
reasonable combination of tolerances that gave me both acceptable values for the 80% EE
and distortion. Further, after consulting with the IAG, I determined these tolerances were
within the ability of the UF machine shop and most major manufacturers. The table with
the output from CODE V using the final tolerances is shown in Table 2-2. I present the
tolerances themselves in Table 2-3.
Once the tolerancing was complete, I was ready to begin the opto-mechanical design.
This process and the outcome is described in the next chapter.
47
Table 2-1. Values of 80% Enclosed Energy for initial tolerancing run with α = β = γ =0.0001 degrees for all optics
X a Y a Nominal Value b 50% b 84.1% b 97.7% b 99.9% b
0.0000 -0.0285 26.4 29.2 35.2 41.3 47.40.0000 0.0000 25.0 30.7 36.3 41.9 47.50.0000 0.0285 22.2 35.0 41.2 47.5 53.80.0000 -0.0140 21.7 29.1 33.5 37.8 42.20.0000 0.0140 24.9 32.1 37.7 43.3 48.90.0140 0.0000 22.8 30.7 35.4 40.2 44.90.0285 0.0000 29.1 34.5 40.2 46.0 51.8
-0.0140 0.0000 22.8 30.7 35.5 40.2 45.0-0.0285 0.0000 29.1 34.5 40.3 46.1 51.9a units are degreesb units are µm
48
Table 2-2. Values of 80% Enclosed Energy for final tolerancing run
X a Y a Nominal Value b 50% b 84.1% b 97.7% b 99.9% b
0.0000 -0.0285 26.4 26.6 29.6 32,7 35.80.0000 0.0000 25.0 25.7 28.4 31.1 33.80.0000 0.0285 22.2 25.2 26.7 28.2 29.70.0000 -0.0140 21.7 23.8 25.8 27.8 29.80.0000 0.0140 24.9 25.8 28.2 30.7 33.10.0140 0.0000 22.8 24.3 26.1 27.8 29.50.0285 0.0000 29.1 29.6 31.9 34.3 36.6
-0.0140 0.0000 22.8 24.4 26.1 27.9 29.6-0.0285 0.0000 29.1 29.6 32.0 34.4 36.7a units are degreesb units are µm
49
Table 2-3. Final tolerances for decenters and tilts. These were the most sensitive (andtherefore most important) values in my tolerancing analysis. Tolerances for theX, Y, Z decenters and α, β, γ tilts are not total errors for all three axes, but foreach axis individually.
Mirror Curvature Decenter (X,Y,Z) Tilt (α,β,γ)% error mm (◦)
C1 0.15 0.037 0.0003C2 0.20 0.037 0.0003Im1 0.09 0.037 0.0003Im2 0.50 0.037 0.0003Im3 0.15 0.037 0.0003Im4 0.10 0.037 0.0001
50
Figure 2-1. The initial 8-lens optical design of CIRCE.
51
Figure 2-2. ORA’s preliminary 6-mirror optical design of CIRCE.
52
Figure 2-3. The standard right-handed Cartesian coordinate system used by ZEMAX andCODE V. Note the X-axis is positive into the page and negative out of thepage.
53
Figure 2-4. An example of the ZEMAX user interface spreadsheet.
54
Figure 2-5. The relation between radius of curvature (RC) and mirror size. For a sphericalmirror, the radius of curvature is the distance between the center of the“sphere” from which the mirror was cut (C) and the vertex (V). Thus thelarger the radius, the larger the optic. A) An optic approximately the size ofthe the beam would have the wrong radius of curvature to direct the light tothe next optic. B) A small segment of a larger parent mirrors allows thedesigner to choose the radius of curvature necessary to correctly direct thelight.
55
Figure 2-6. Parent mirrors used in the preliminary CIRCE design pictured in Figure 2-1.Note that the parent mirrors are much larger than the beam size and oftencollide with one another. Using smaller segments of these mirrors greatlyreduces cost, makes manufacture more practical, and allows light to propagatethrough the system.
56
Figure 2-7. The original Folded Cassegrain focal station envelope. Note the 1000 mmdiameter of the cylinder. The 2000 mm length includes acquisition and guideand rotator hardware; the length specified for CIRCE was only 1500mm.
57
Figure 2-8. The University of Florida CIRCE redesign. The addition of the foldscompressed the design along the Y-axis and allowed it to fit inside the GTCFolded Cassegrain envelope. A) A layout of CIRCE with X, Y, Z = 0. B) Thelayout rotated to Y = 270 degrees. Note that this configuration called for twooptical benches or a single bench with optics attached to both sides.
58
Figure 2-9. The final CIRCE optical design. To fit into the GTC envelope, ORA addedtwo fold mirrors, F1 and F2, before the first collimator, C1, and rotated thefirst imaging mirror, IM1, to direct the light in a positive Y direction. In thisdesign, light passes through the entrance window (1), to the telescope focalplane (2), before being reflected by two fold mirrors into the collimating optics(C1 and C2). The collimated light propagates to the Lyot or cold stop (3) andpasses through a filter (4), before being imaged on the detector focal plane (5)by the four-mirror camera.
59
Figure 2-10. The footprint at the detector focal plane. The square aperture defines thearea of the HAWAII 2K infrared detector that we will place at the focalplane. The marks (circled for easier identification) represent rays launchedfrom 8 locations that mark the boundaries of a 3.4′ × 3.4′ area on the “sky”.This defines our FOV; any positions on the sky within this FOV will alsoland on the detector.
60
Figure 2-11. The J-band encircled energy diagram. I chose five positions on the detectorto measure the amount of energy enclosed as a function of distance. Thesefive positions are the center (0(◦,0(◦), the bottom left corner (-0.0280◦,-0.0280◦), a spot halfway (0.014◦, 0.014◦) and 3/4 (0.025◦, 0.025◦) betweenthe center and the upper right corner, and a position 2/3 between the centerand bottom right corner. Note that the EE curves for each positions arealmost identical to positions in different quadrants but at similar radii fromthe detector. The black curve represents the diffraction limited case, the bestachievable EE.
61
Figure 2-12. Same as Figure 2-9 but for the K-band. Note that the EE enclosed within aradius is slightly lower for every position on the detector. Still, we findexcellent values; the EE in two pixels at the very corner of the detector is >70%.
62
Figure 2-13. The K-band spot diagram for CIRCE. We note that for many field positions,the majority of the rays fall within one Airy Disk (shown by the black circle)and are tightly packed. The corners show the most aberration, but no oneparticular type dominates and the spot sizes are still small, if irregularlyshaped.
63
CHAPTER 3OPTO-MECHANICAL DESIGN OF CIRCE
3.1 Basics of Mechanical Design
In Chapter 2, I reviewed the optical design for CIRCE, detailing how I [1] used the
instrument’s science drivers to constrain broad optical characteristics, [2] helped create an
optical layout that met these specifications, [3] analyzed the expected performance, and [4]
evaluated the likely performance, accounting for potential errors during manufacture. The
next step was to complete the CIRCE opto-mechanical design. Opto-mechanical design is
the process of creating mechanical drawings of precision optics and associated hardware
using details from the optical prescription.
Mechanical drawings depict objects to scale and include the precise dimensions
and tolerances necessary to manufacture a desired piece. While the two- and three-
dimensional layouts presented in Chapter 2 describe only the front surface, the aspheric
“face”, of each optic, mechanical drawings show every surface of the mirror. Furthermore,
optical layouts display the mirror with a “dummy” thickness of zero. In reality, mirrors
are thick blocks of 6061-T6 aluminum with a reflective surface on the front and bolt and
pin holes on the back (see Figure 3-1).
The first step in manufacturing a mirror is to create a “blank”, a physical block of
material machined to the length, width, and height of the final optic. Next, a machinist
adds bolt and pin holes to the back surface of the blank. Once this is complete the
aspheric surface described by the optical prescription is diamond-turned onto the blank.
In this process, a Computer Numerical Control (CNC) lathe will rotate the front face of
the blank against a diamond-tipped cutting tool to produce the exact aspheric surface in
the optical prescription. Diamond-turning typically produces sub-µm form accuracy and
sub-nm surface accuracy. After the diamond-turning is complete, the mirror is plated with
a gold coating, which is 97% reflective in the NIR.
64
For manufacturing purposes I needed to create a separate mechanical drawing of
each optic. Besides noting the length, width, and height of every mirror, the mechanical
drawings also provide a reference between the complex aspheric surface on the aluminum
block and a particular pin or bolt hole on the back face of the blank. This is the basis
of the “bolt-and-go” or “bolt-together” technique mentioned in §2.1; alignment and
placement of optics is easier if the reference is a well-defined physical hole on the back of
the mirror instead of the mechanical center of a theoretical aspheric surface.
In the following sections, I will detail my work on the opto-mechanical design of
CIRCE. Specifically, I discuss [1] the steps I took to create two-dimensional mechanical
drawings of each optic and its bracket for manufacturing purposes, [2] opto-mechanical
layouts of the entire system in both two and three dimensions, and [3] a preliminary
design of the optical bench.
3.2 Mechanical Design Software
While one can produce mechanical drawings with nothing more than a pencil,
paper, and ruler, modern technology offers more advanced options; specifically Computer
Assisted Design (CAD) programs that allow users to manipulate lines and shapes to
produce accurate representations of objects. One of the most popular CAD programs
is AutoCAD. Along with its offshoot, Mechanical Desktop, AutoCAD is an industry
standard and used for both two- and three- dimensional mechanical design (Bethune
1997).
AutoCAD’s native geometry is a Cartesian coordinate system with the X-axis
increasing positively to the right, the Y-axis increasing positively to the top, and the
Z-axis increasing positively out of the paper. When drawing in AutoCAD, users may
change their viewpoint and see different faces of an object. However, the definitions of
X, Y, and Z used when entering dimensions for an object do not change regardless of the
object’s orientation.
65
It is useful to point out that while most optical designers and design programs use
metric units, mechanical drawings in the United States are almost always in English units.
This is for practical manufacturing purposes; most lathes, drill presses, and other tools
in the US use English units. Therefore all mechanical drawings discussed in this section
are dimensioned in inches. This is important to be aware of as all the dimensions and
decenters from ZEMAX and Code V, as well as the tolerances necessary for the designs
discussed in §2.7, are in millimeters.
The easiest method for drawing complex shapes is to deconstruct the object into its
basic shape, draw these shapes to the correct scale, and then use tools within AutoCAD
to alter them until the result replicates the desired object exactly (Bethune 1997). For
objects created from scratch, this process requires that the designer has either a very
specific list of dimensions for the final object from another source, or firsthand knowledge
of the object’s specifications and purpose. For CIRCE, the latter situation applied (see
§3.3). Either option allows the designer to create the most accurate drawing possible for
the manufacturer.
While AutoCAD allows users to model objects in both two- and three- dimensions,
it is easer to label and less confusing to read two-dimensional drawings. Thus, most
mechanical designs sent to manufacturers are presented in two-dimensions. Two-dimensional
drawings show a slice or cross-section of a three-dimensional object. Usually this is
accomplished by taking a plane and slicing an object perpendicular to one axis as in
Figure 3-2. For instance, an XY cross-section requires slicing perpendicular to the Z-axis;
in effect, holding it constant at one value. This allows us to see the shape of an object at a
particular value of Z.
To create a cross-section, designers have two options. For simple objects, where the
specifications of the cross-section are well-known, they may draw the cross-section directly
using two-dimensional shapes. If the object is very complex and the exact dimensions
of its cross-section difficult to extrapolate, a designer might opt instead to draw the
66
object in three-dimensions and use AutoCAD to produce slices of the appropriate planes.
Three-dimensional drawing, also called solid modeling, uses shapes like spheres, cylinders,
cones, etc. to construct an exact model of an object.
I employed both two-dimensional drawing and solid modeling to produce mechanical
drawings of the CIRCE optics. I then used these designs to determine the positions
of the mirrors on the optical bench, – a flat, table-like surface that supports brackets,
optics, cryo-mechanisms, electronics and the detector and create brackets with the
correct dimensions and hole patterns to hold them in place. I will first discuss the initial
two-dimensional designs of the mirrors and the process that I used to determine the
correct dimensions for the blanks.
3.3 Designing and Drawing the CIRCE Mirrors
As discussed above, designers can create two-dimensional drawings in one of two
ways: directly with two-dimensional shapes or indirectly from solid models. Although
solid modeling is very powerful (see §3.6), it is also more time-consuming. Since my first
drawings were primarily used to consult with members of the InfraRed Astrophysics
Group (IAG) regarding locations and sizes of bolt and pin holes, “blank” dimensions, and
mechanical drawing techniques, I initially used the faster 2-D method. This allowed me to
make corrections very quickly, reducing the time spent on each iteration.
Before beginning my drawings, I needed to determine the physical size of each mirror.
Since we specified that CIRCE should have a 3.4′ × 3.4′ FOV, (§2.1) any light from
within this area of sky should fall completely on every optic. Thus, two dimensions of each
mirror, the width and height, were linked to the optical design. If a mirror is undersized
in either dimension, light at the edge of the field will miss the mirror, thus failing to
propagate to the next optic.
To avoid any loss of photons, I ensured that the mirrors were large enough to capture
and reflect light from the entire field. In ZEMAX, I launched beams from ten field angles,
marking out the FOV of CIRCE. Using footprint diagrams (see §2.5) I mapped where the
67
beams landed on each mirror. I enlarged the width and height of the mirrors one-by-one
until the footprint diagram for every surface showed that all the beams fell completely on
the mirror. I then added 6mm to the height and width. This created a 3mm buffer around
the edge of each mirror, where defects and tool marks are most likely to occur during
machining. This prevents unplanned aberrations caused by the manufacture process. I
present the footprint diagrams of the eight CIRCE mirrors with corrected apertures in
Figure 3-3. Table 3-1 gives the final widths and heights of the mirrors. Recall from §2.2
that X corresponds to height, Y to length, and Z to width in optical design parlance. It is
important for later discussions §3.8 to recognize that many of these optics are quite large;
the largest mirror, Collimator 1, is 10 inches × 10 inches × 5 inches.
With the dimensions of the optics now well-defined, I produced sketches of the
YZ and XY cross-sections of each mirror. Here, I present and discuss in detail the
mechanical drawings created for the first collimating mirror; however I also completed
similar drawings for the remaining seven optics.
Figure 3-4 is an XY cross-section of collimator 1. This slice shows the mirror as it
would look if we viewed it from its back surface. It gives both the length and height of the
mirror, as defined by the footprint analysis, and defines the bolt- and pin- hole patterns on
the back surface of the optic. The two pinholes and three bolt holes will secure the mirror
to its bracket.
Pins are tight fitting cylinders of metal that are used to precisely position optics. The
locations of the pinholes are constrained by tight tolerances, generally in the range of 10
- 25 µm in each direction. Since the locations of the pinholes are only one possible source
of error when positioning and aligning our optics, the tolerance for each hole must be less
than the total 37 µm budgeted in our tolerancing analysis for each of the the X, Y, and Z
decenters.
The pinholes are also referenced very accurately to the aspheric surface of the mirror.
Two pinholes per optic are necessary: one to fix the lateral motion and the other to
68
prevent rotation. Three bolt holes per optic lie at 120 degree intervals on a circle centered
on the central pinhole. The bolts support the bulk of the mirror’s weight and secure it
to its bracket. However, the front of the bracket and back of the mirror do not sit flush
against each other; instead three raised “ contact pads” surrounding the bolt holes are
the contact points between the bracket and mirror. These three points of contact provide
the most stability and flatness when mounting the mirror. This is a geometric argument -
since three points define a plane, any object with three contact points will stabilize itself
on a flat surface.
In Figure 3-4, I indicate several dimensions, including the size of the contact
pads, bolt holes, pinholes, and bolt circle. I used mirror drawings from a previous UF
instrument, T-ReCS (Telesco et al. 1998), as a general reference for screw and pin sizes.
As the CIRCE design progressed, recommendations from the IAG led me to alter many of
these specifications. I discuss this process in more detail.
The next two dimensional drawing, Figure 3-5, is a YZ cross-section, sliced at the X
= 0 plane. To make the YZ cross-section I considered two surfaces, the back and front
of the mirror. Both surfaces had the same Y dimension, or width, set by the footprint
diagram analysis outlined above. The front surface of the mirror was modeled using
Equation 2–1. Since this aspheric equation defines Z is a function of Y, we chose critical
values for Y at the top, bottom, and mechanical center of the mirrors; computed the
corresponding values of Z; marked them in AutoCAD, and connected them with a curve.
Recalling that the mirrors in the CIRCE design are segments of larger parent mirrors (see
§2.2, we also marked the vertex, the center of the parent mirror where Y and Z = 0, and
the aperture decenter, or distance from the mechanical center to the vertex.
I then diagrammed the back surface of the mirror, as seen in the YZ cross-section,
referencing Figure 3-4 for positions and sizes. Using the locations and depths of the bolt
and pin holes and contact pads specified on the XY cross-section, I modeled these objects
with two-dimensional shapes available in AutoCAD.
69
Examining Figure 3-5, we see that the width of the mirror (∆ Z) is not constant.
This is because Z is a function of Y on the front surface of the optic but constant on the
back, creating a uniform back to mount to a flat bracket. To find the correct widths for
the mirrors, I consulted mirror designs from the previously constructed UF instrument
T-ReCS (Telesco et al. 1998). I found that the minimum width of each T-ReCS mirrors
was ∼ 25-30% of its length, which guaranteed rigidity, preventing the optic from bending
either during manufacture or once mounted in the instrument. This rule-of-thumb for
sizing optics was confirmed by IAG engineers and I applied it to all the CIRCE mirrors.
For example, the length of the Collimator 1, is ∼ 10 inches, so the minimum width, as
shown in Figure 3-5, is ∼ 2.7 inches.
I completed two-dimensional drawings of the XY and YZ cross-section of all six
powered CIRCE mirrors following the procedure described above. I created similar
drawings for the two unpowered fold mirrors; with only minor differences in the YZ
cross-sections. The final dimensions for all optics are listed in Table 3-1. Finally, I
combined the XY and YZ cross-sections for each optic into one mechanical drawing for
the manufacturer’s use. I show an example of the completed diagram of Collimator 1 in
Figure 3-6.
With the first two-dimensional sketch of the CIRCE optics complete, I presented my
work to the IAG for review. Senior UF mechanical engineer Jeff Julian advised me to
make several changes to the drawings including the addition of the aspheric equation on
each aspheric mirror drawing; a list of important notes for manufacture regarding surface
irregularities, coatings, and clear aperture (determined from the footprint of the beams;
see §3.3); more detail regarding the locations of the pin and bolt holes; and instructions
requiring specific surface flatness to make sure the contact pads were smooth.
Further, the IAG suggested important alterations to the mirrors themselves. First,
they recommended that I add bolt “reliefs” to alleviate pressure from the mounting
bolts. Reliefs are semi-circular cut-outs toward the back of the mirror that intersect the
70
bolt holes near their greatest depth. Since many of the mirrors are quite large (recall
Collimator 1 is 10 inches × 10 inches × 5 inches), the weight of the optic pulling on
the bolt could slightly warp the surface of the mirror opposite the hole. The reliefs
would alleviate this pressure and prevent warping. I implemented this correction to the
mirrors and made the suggested additions to the drawings. I show an example of a new
mechanical design with a close-up of the reliefs in Figure 3-7.
The IAG also suggested a modification to the fold mirrors. As shown in Figure 3-8,
the fold mirrors were initially rectangular blocks. To minimize the total weight of the
instrument and reduce the area the folds occupy on the bench, I removed material from
the sides of the mirror so that the sides tapered from the front to the back surface. I show
an example of a new fold trapezoidal-shaped fold1 mirror in Figure 3-9.
3.4 Opto-Mechanical Layout
In §3.3, I discussed the creation of individual mechanical drawings for every mirror
in the CIRCE optical design. In the remainder of this chapter, I present my work on the
opto-mechanical layout, which is a mechanical design of the complete optical assembly
including the mirrors, brackets, and the optical bench.
As discussed in the introduction of this chapter, optical software programs like
ZEMAX only model the front surface of each CIRCE mirror in the optical layout. These
surfaces are ultimately machined onto 1 - 5 inch thick blocks of aluminum (see §3.3) with
heights and widths of up to 10 inches. Thus, while the ZEMAX optical layout portrays
the aspheric surfaces of CIRCE with precision, it does not provide a realistic picture of
what the assembled optical system will look like once it is complete.
The challenge of producing an opto-mechanical layout is to create mechanical
drawings for manufacture that accurately portray how the finished assembly of mirrors
and brackets will appear, while staying true to the optical prescription. The opto-mechanical
layout must model the aspheric surfaces, with the correct locations, decenters, and tilts
71
(see §2.2), as accurately as the optical layout, while showing the precise position of pin
and bolt holes on the back of mirrors and exact locations of the brackets.
Although this sounds simple in principle, the execution is quite difficult. While
optical designers can use ZEMAX to theoretically“place” free-floating mirrors in the
correct position on the optical layout every time, the reality requires more physical detail.
In the assembled instrument, the mirrors will bolt to brackets, which in turn will bolt to
the optical bench. If any of these pieces is designed or machined incorrectly, the aspheric
surface will not be in the correct position with respect to the other optics, adversely
affecting image quality.
In §2.7, I placed limits on the amount of error we could tolerate before the image
quality degraded to an unacceptable level. For instance, the tolerances for the decenters,
the X, Y, and Z position of an aspheric surface with respect to the the previous surface,
are ∼ 40µm in each direction (see Table 2-3). This is the total allowable error for pin and
bolt holes in the bench, brackets, and mirrors for both design and manufacture. While
designing to this level of precision is feasible, it still requires careful work on the part of
the mechanical designer and (as I discovered) powerful CAD tools.
In the following sections, I detail attempts to construct a mechanical design of the
cryo-mechanical layout using two-dimensional cross-sections and three-dimensional solid
models.
3.5 Two-Dimensional Opto-Mechanical Layout
Using the YZ cross-section of the eight CIRCE optics (see §3.3), I created a
two-dimensional opto-mechanical layout of CIRCE. Figure 3-10, shows the first results: a
YZ cross-section of the optical assembly at X = 0.
To create this layout, I first needed the positions and rotations of the mirrors from
the optical prescription. In ZEMAX, I defined a “global reference”, a surface designated as
the reference point by the user. I chose the entrance window, the first optic in the CIRCE
optical path. ZEMAX output the X, Y, and Z location of the vertex of each mirror, as
72
well as the α, β, and γ rotations around the vertex, with respect to the entrance window.
In AutoCAD, I defined the location of the entrance window as X = Y = Z = 0, so that all
positions and rotations would be with respect to the origin.
Since the CIRCE mirrors are off-axis segments of larger parent mirrors (§2.2), their
vertices are not located within the defined aperture. However, the two-dimensional YZ
cross-sections of each mirror, like that shown in Figure 3-5, include the vertex. I imported
all eight YZ cross-sections with their vertices into a single file. Then using a various tools
in AutoCad, I positioned and tilted each mirror by the prescribed amounts.
Once I produced the layout shown in Figure 3-10, I checked to ensure that none of
the mirrors blocked or intersected any other optic. We noted that the GTC re-design
necessitated by the envelope specifications (see §2.3) left very tight spaces between the
first collimator (C1) and fourth imager (Im4) and the first fold mirror (F1) and the second
collimator (C2). Throughout the opto-mechanical design it was critical to be conscious of
potential space issues when placing mechanical elements, like brackets, into the system.
However, as I described in §3.4, the true test of the opto-mechanical layout is
whether the mirror surfaces are in the same location (within the tolerances) as the
aspheric surfaces in the optical design. This is counterintuitive at first; one would expect
a powerful program like AutoCAD to very precisely position any shape. However, given
the complexity of the aspheric surface and its rotation about a distant point (the vertex),
accuracy is not ensured. In fact, when I compared the X. Y, and Z global position of the
mechanical center derived from ZEMAX, I found that the AutoCAD layout was off by
over 100 µm in Y and 1000 µm in Z for collimator 2. Further, two of the imaging mirrors,
two and four, had errors > 20 µm in more than one direction. While these errors were
below our ∼ 40 µm tolerances, if we accepted the AutoCAD layout as-is, this level of error
would be incorporated into the design, leaving little room for manufacturing or placement
errors for these mirrors later on.
73
After several attempts to track down possible sources of error in the software with no
improvement of the AutoCAD positions, the CIRCE team decided to move on to a more
robust method: solid modeling. However, this work ultimately proved useful; I determined
that the CIRCE optics would fit into the space provided and developed a technique for
comparing the global location of any point on a mirrors in AutoCAD and Zemax.
3.6 Design and Analysis of Three-Dimensional Mechanical Layout
Seeking a solution to the layout problem, I explored solid modeling in AutoCAD.
Solid modeling is the creation of three-dimensional models using spheres, boxes, cylinders,
cones, and other solid shapes. Using AutoCAD, a designer can slice models to create
cross-sections, rotate models for views of any face, and add and subtract solids to build
more complex objects. Solid modeling is often more intuitive then two-dimensional
modeling; one creates an object exactly as one would see it in reality, as opposed to
imagining cross-sectional views or slices in perfect detail.
Since AutoCAD provides no method to produce aspheric surfaces from scratch, I
made simplistic solid models of the CIRCE optics using spherical surfaces (see Figure 3-11
for an example). Using a similar technique to the one developed for the two-dimensional
layout, I placed the solid models on an optical bench with the correct decenters and tilts.
Although this layout had the same basic problems as the previous design, exacerbated by
the fact that the shape of CIRCE’s powered optics are not spherical, I could, for the first
time, probe the CIRCE optical mechanical design from every angle, checking that mirrors
did not intersect each other off the X = 0 axis; this was not possible to do before because
of the limited viewpoint offered by the two-dimensional design.
I then sought to improve this design and complete the three-dimensional opto-mechanical
layout by modeling the mirrors with the correct aspheric surface. While ZEMAX can
export surfaces and beams to files readable by AutoCAD, the resulting three-dimensional
design was so complex, with thousands of objects and intersecting surfaces, editing the file
was too difficult and time-consuming to be practical.
74
Instead, working in collaboration with Senior Mechanical Engineer Jeff Julian, we
iterated several times with various ZEMAX outputs and two dimensional designs until he
produced an accurate three-dimensional optical layout of CIRCE in SolidWorks, another
powerful mechanical design program. This resulting design, presented in Figure 3-12, is
also readable by AutoCAD.
Applying the technique described in §3.5, I analyzed this design, using the four
corners of the mirrors’ front surfaces to compare positions in AutoCAD against positions
in ZEMAX. The results were excellent; I found an average difference of < 1 µm in X, ∼ 5
µm in Y, and ∼ 3 µm in Z. The largest difference along any axis was ∼ 15 µm.
Using my two-dimensional drawings as a reference, I checked the locations of bolt
holes, pinholes, contact pads, and bolt reliefs making additions and changes where
necessary. Then, using this solid model, I completed two-dimensional cuts of all the optics,
creating XY and YZ cross-sections of each optics for mechanical drawings.
3.7 Development in the Manufacture of the CIRCE Optics
During the opto-mechanical design of the CIRCE mirrors, the CIRCE team contacted
several vendors to find a manufacturer for the optics. After consulting with several
companies, we noted that the price for the basic machining of the blanks and subsequent
testing of each optic was approximately half the total cost of the optics.
To drastically reduce costs, Janos Technology, one of the potential vendors, and
the CIRCE PI developed an alternate plan. Rather than having Janos manufacture the
blanks, the Astronomy and Physics Machine Shop at UF would machine them. The
UF shop would also produce the brackets and the optical bench. These pieces would be
shipped en masse to engineers at Janos, who would diamond-turn the mirrors, set-up the
entire bench, and test the optics in concert, rather than individually testing each optic.
Any aberrations caused by errors in the decenter, tilt, or mirror curvature would be found
using inteferometry and corrected by slightly altering the shape of the mirrors. Then
75
Janos would gold coat the mirrors, re-mount the optics, and send the entire optical bench
back to UF, where we could test and install it directly into the cryogenic dewar.
The tolerances that I calculated would still be useful guidelines for the mechanical
design of the optical bench and the precision machining of the pin and bolt holes on each
mirror. However, with this process we would avoid the majority of decenter and tilt errors,
which generally surface during the alignment process.
The CIRCE PI consulted with the CIRCE team and we decided that this was an
elegant and frugal plan. The only potential downside was the extended wait; while the
design of the optics was nearing completion, design of the brackets and bench had not
yet begun. Janos could not start work on the optics until we delivered the mirror blanks,
brackets, and bench.
In the meantime, using the mechanical drawings produced in §3.6, the Physics and
Astronomy Machine Shop at UF produced tests of all eight CIRCE optics. Two of these
optics, the second and third imager are shown in Figure 3-13. After examining these test
mirrors, we noted that the weight of the larger optics was problematic. For instance,
Collimator 1, the largest mirror, weighed > 23 kgs. Considering weight constraints set by
the GTC (∼ 800 kgs total for the whole instrument) and concerns that the mirror would
be difficult to support with a bracket, we decided to lightweight the larger CIRCE optics.
3.8 Lightweighting the CIRCE Optics
The first step in lightweighting the CIRCE optics was to decide which mirrors needed
modification. After carefully considering our optics, we chose both flat mirrors, the first
collimator, and the first and fourth imager. These optics are twice the size of the smaller
optics and estimated to weigh over 14 kgs each.
Generally when lightweighting optics, two options are available. The first is to
manufacture the mirror blank from a light material. For instance, the secondary mirror
of the GTC is made of beryllium (Devaney et al. 2004), a low density material that is
also very strong. However this solution is unsuitable here because these materials [1]
76
are prohibitively expensive for the CIRCE budget and [2] have coefficients of thermal
expansion different than 6061-T6, nullifying one of the major benefits of the all-reflective
design,namely homologous contraction.
The second choice was to remove large blocks of material from each mirror blank
without compromising the rigidity of the optics. A common technique that meets this
criterion is to create a “honeycomb” pattern on the back surface of the optic. Honeycomb
patterns are both very strong and capable of greatly reducing the weight of an optic; one
honeycombed 8.4-meter mirror on the Large Binocular Telescope is an order of magnitude
stiffer and 40% lighter than a similar sized solid blank (Hill et al. 1998). Honeycomb
patterns are also used to lightweight other opto-mechanical elements, such as optical
benches.
We decided to employ a slightly modified version of this method, removing triangular
pieces of material of varying sizes and depths to create honeycomb-like hexagonal
structures on the back of the mirrors. This was easier, and thus less expensive, than
machining many small intersecting hexagons.
I first created a two-dimensional sketch of a honeycomb pattern for each of the five
larger mirrors, avoiding pin and bolt holes while maximizing the number of triangles on
the back surface of the optic. I show a cross-section of one of these patterns in Figure
3-14. Next I used this pattern to construct the three-dimensional honeycomb structure.
I calculated the depth of each triangular pockets, taking into consideration the IAG
recommendation of leaving at least 1 inch from the front surface of the mirrors intact.
This precaution would prevent any bulging or pressure on the mirror surface from the
tools used to create the pockets. Since the mirror depth varied across the Y axis, the
depth of the pockets also varied; the ones at the thicker part of the mirror were deeper
than those at the slimmer end.
I then removed the bolt reliefs added in an earlier phase of the mirror design. This
decision was based on two considerations. First, lightweighting the mirrors decreased the
77
need for reliefs; the pressure on the bolts that the reliefs alleviated was much smaller due
to the mirror’s lighter weight. Secondly, the reliefs inteferred with the new lightweighting,
unavoidably driving through the now honeycombed pockets, which threatened to weaken
the entire structure. Since the lightweighting was crucial and lessened the need for the
relief holes anyway, we chose to remove the latter.
An example of a finished three-dimensional lightweighted mirror is shown in Figure
3-15. Analysis of the mirror designs using AutoCAD estimation tools showed that we
reduced the weight of the mirrors by 20 - 25% with our lightweighting technique.
3.9 Finalized Mechanical Drawings of the CIRCE Mirrors
With our three-dimensional designs of lightweight mirrors complete, we moved on
to produce finalized two-dimensional mechanical drawings. We completed two sets of
these drawings, one for the machine shop at UF and one for Janos. The set of drawings
for the mirror blanks were slightly modified; we removed the aspheric surface in the YZ
cross-section and replaced it with a flat surface with slightly larger dimensions than the
final mirrors. This ensured that Janos would have enough material to diamond-turn the
correct aspheric surface. I present an example of the blank drawings in Figure 3-16. I
show an example of the accurate two-dimensional mechanical drawings prepared for Janos
from the lightweighted solid models described in §3.8 in Figure 3-17.
3.10 Mechanical Design of the Mirror Brackets
Upon completing the two- and three- dimensional design of the CIRCE mirrors, I
began to design the mirror brackets. Brackets are machined pieces of 6061-T6 aluminum
that attach to both the back of the mirror and top of the optical bench. They serve to
suspend the mirror above the bench so that the optical axis of the system intersects the
mirror at the correct height. They also firmly attach the mirror to the liquid nitrogen
cooled optical bench.
While a variety of bracket and brace types exist, the CIRCE team took the advice of
the IAG and chose the strong and simple to manufacture “L” and “T” shaped brackets.
78
Figure 3-18 shows an example of both a simple L- and T- bracket. For CIRCE, we
also added gussets; these are triangular shaped wedges that add rigidity to the vertical
member. I discuss this in greater detail later in this section.
Before I could begin to manufacture the brackets, I determined how high off the
bench the mirrors would sit. CIRCE is designed so that the mechanical center of every
optic is aligned with the X = 0 axis, or “optical axis”. The optical bench will therefore
rest at a negative X value in the final three-dimensional design. The exact distance
between the optical axis and the bench depended on the largest optic: the first collimator
. The height of the Collimator 1 is 10.037 inches; thus, the bench needs to be at least
5.0185 inches below the X = 0 axis. We added extra space for clearance, so that the
mirror would not rest directly on the bench. In the end the optical bench was inserted
into the optical design at X = -5.375 inches.
With this information, I could design the brackets. First I determined the size of
the upright part of the bracket, the piece that attaches to the mirror. The width of
this support, the Y-dimension, was the same as the Y-dimension of its mirror. The
X-dimension or height of the upright was equal to the X-dimension of its mirror added
to the mirror’s height off the bench. For instance, a mirror with a height of 6.000 inches,
centered on X = 0 would extend down to X = -3.000. Since the bench rests at X =
-5.375, the distance between the bottom of the mirror and the bench is 2.375. The upright
would therefore be 8.875 inches: 6.000 inches added to 2.375 inches. Finally, the length or
Z-dimension of the bracket was 0.5 inches, rigid enough to resist flexure due to the weight
of the optic without adding significant extra weight to each bracket.
We then designed the horizontal base of the L and T brackets. In both cases, the
Y-dimension of the bracket was set by the Y-dimension of the mirror. The X-dimension,
or height, was set as 0.5 inches, as discussed above. The Z-dimension was set by
the height of the upright. For L brackets, the IAG recommended that the base be
approximately 50% of the height of the upright. T brackets were less constrained; I
79
chose to make the horizontal base the same size as the upright, so that either side of the T
base met the same 50% rule-of-thumb.
With this information, I completed three-dimensional solid models of the CIRCE
brackets. After drawing the basic shapes that composed a bracket, I added pinholes to the
base. The pinholes served the same purpose as the pinholes for the mirror; they secured
the bracket in a precise location. Again, the tolerances on these pinholes were very tight;
∼ 20 µm in each axis. The size of the hole was specified to 0.0005 inch or ∼ 10 µm. I then
added bolt holes that we will use to tightly secure the bracket to the optical bench.
I then designed the gussets (wedges that support and strengthen the upright). In the
preliminary designs, the wedge-shaped gussets for the larger mirrors bisected the bracket,
starting at the middle of the upright, as shown in Figure 3-18. However, the CIRCE team
decided that instead of one thick gusset, two thinner gussets at the edges of brackets
would distribute the weight more effectively and provide a stiffer support (see Figure 3-19)
Meanwhile, we designed smaller brackets with three gussets, two in the front at the edges
and one in the back bisecting the bracket. I present an example of this type of bracket in
Figure 3-20.
After completing the basic solid models of the brackets, I imported them into our
three-dimensional layout of the CIRCE optics. Using a variety of tools in AutoCAD,
I aligned each bracket precisely to its mirror. This required making sure that only the
contact pads on the mirrors touched the bracket. I then copied the bolt- and pin- hole
patterns directly from the back of the optics onto the brackets, ensuring that that they
matched.
Next, I checked for any places in the layout where a bracket intersected another
bracket or mirror. I noted that the back gusset on the bracket for the second collimator
mirror blocked the optical path of light traveling from the first fold mirror to the second
fold mirror. I reduced the size of the gusset, solving the problem. The first collimator
and fourth imager presented a more serious problem. Both mirrors are very large and
80
require equally large brackets. However, examining Figure 3-11 we see that there is little
space between the two optics. The solution was to make a compound bracket, a large base
with two uprights. Instead of a gusset, a removable wedge supported both vertical pieces.
This allowed room to tighten bolt holes with the necessary tools while providing adequate
rigidity. Figure 3-21 shows the final solid model of the compound bracket.
I then created two-dimensional drawings of the brackets for manufacture. These
drawings featured three cross-sections (XY, YZ, XZ) and a solid model. I present an
example of these drawings in Figure 3-22. In Figure 3-23, I show a picture of two of the
finished brackets, manufactured by the UF Physics and Astronomy Machine shop. The
brackets are bolted to their respective “practice” mirrors, which were discussed in §3.7
3.11 Preliminary Designs of the Optical Bench
The optical bench that supports the optical components of CIRCE is located directly
above a tank of cryogen, in this case liquid nitrogen, the temperature of the bench is ∼ 77
K. All components must have excellent thermal contact with the bench to remain cold.
The size of the bench determines the overall size of the dewar, as the heat shields
and the vacuum jacket are generally positioned at known distances from each other in
relation to the bench. In turn, the bench size is set by the optics. As indicated in §2.3
the CIRCE optics are centered at Y = 0. The farthest optical element from this central
position is the first collimator at Y ∼ 17 inches. Allowing another 1 inch from the edge for
clearance, yields a bench width of 36 inches. We then found the furthest optical element
in the Z-dimension, the detector focal plane at Z ∼ 57.5 inches. Adding another 3 inches
to account for the electronics necessary to run the detector and the detector stage, we
calculated a Z-dimension for the bench of 60.5 inches.
The thickness of the bench is of critical importance. If the bench is too thin, it
will warp under the weight of the optics and cryogen. This warping or bending is called
flexure. If the flexure is too great, the instrument will bend and the positions of individual
mirrors will change as the telescope and attached instrument moves. Thus, light from
81
the telescope will not propagate correctly, striking optics in the wrong position on their
surface, creating aberrations in the system.
Miguel Charcos, a CIRCE team member, created a program in IDL to measure
flexure in the bench. Considering a cantilever beam (a horizonal member suspended on
one side from a vertical upright) he calculated the flexure of the bench given its thickness
and the total weight of the mirrors, bench, and liquid nitrogen. He found that for a bench
with a thickness of 2 inches, the flexure was less than 20 µm, the same order of magnitude
as the tolerances on our X, Y, and Z, decenters. Further, this was the “worst-case”
scenario, calculated with all the optics at the very edge of the bench.
Using this bench thickness, I constructed a preliminary CIRCE optical bench. To
add rigidity and reduce the weight, I applied the same honeycomb pattern as used on the
CIRCE mirrors to the underside of the bench. I then positioned this bench at X = -5.375,
as discussed in §3.10. Using tools in AutoCAD, I mirrored the pin and bolt hole patterns
from the brackets onto the bench to ensure that they matched. I present a solid model of
the bench in Figure 3-24 and the completed three-dimensional opto-mechanical design of
CIRCE in Figure 3-25.
82
Table 3-1. Dimensions of the CIRCE mirrors. ∆X corresponds to height, ∆Y to length,∆Z min to the thinnest width, and ∆Z max to the thickest width.
Mirror ∆X ∆Y ∆Z min ∆Z max(in) (in) (in) (in)
Fold 1 7.480 9.370 1.980 1.980Fold 2 8.189 10.394 1.980 1.980Collimator 1 10.079 10.079 2.020 3.965Collimator 2 4.488 4.567 1.200 1.367Imager 1 7.244 6.772 1.820 5.413Imager 2 3.071 3.780 1.000 1.352Imager 3 2.913 3.386 0.900 1.123Imager 4 7.402 8.189 1.720 5.055
83
Figure 3-1. The difference between optical and opto-mechanical representations of ahypothetical spherical mirror. A) Optical programs like ZEMAX only modelthe front surface of the mirror B) Using mechanical drawings I can create amore detailed image of the entire mirror, with hole patterns and a realisticthickness.
84
Figure 3-2. The relationship between a mechanical cross-section and a three-dimensionalsolid model. A) The solid model of an optic bisected by a plane at X = 0. B)Removing the top half of the mirror – the purple surface represents theintersection of the plane with the optic. C) Removing the entirethree-dimensional optic, we are now left with only the intersection of themirror with the cutting plane, in this case, a YZ cross-section.
85
Figure 3-3. K-band footprint diagrams of the CIRCE Mirrors. Note that the beams werelaunched from nine field positions outlining a 3.4′ × 3.4′ area of sky. Allmirrors were sized to capture all beams from these positions and thenoversized by 6 mm in each dimensions to allow for machining defects, whichare most common around the edges of optics.
86
Figure 3-4. Preliminary XY Cross-section of collimator 1. The dimensions of the mirror,locations of the pin and bolt holes, radius of the bolt hole circle and positionof the contact pads are included on the diagram.
87
Figure 3-5. Preliminary YZ Cross-section of Collimator 1. The drawing shows the asphericsurface, the distance to the vertex, and the thickness of the mirror at theedges.
88
Figure 3-6. Preliminary two-dimensional mechanical drawing for Collimator 1 showingboth XY and YZ cross-sections. Note that the section labeled A-A shows theYZ cross-section for a cut taken across the line labeled A-A in the XYdrawing. This line represents the X = 0 plane, which bisects all the CIRCEoptics.
89
Figure 3-7. Revised two-dimensional cross-sections of Collimator 1 following a designreview. Note the addition of bolt reliefs. A detail of this feature is shown thecircle labeled B.
90
Figure 3-8. The original 2-D mechanical drawing of Fold 1. The block shape of the mirrorwas revised to reduce weight. See Figure 3-9 for the new design of Fold Mirror1.
91
Figure 3-9. Fold Mirror 1 after a design review and redesign. The sides now slope inward,which reduces both the mirror’s weight and the space they occupy on thebench.
92
Figure 3-10. The 2-D opto-mechanical layout of CIRCE. This shows the first attempt toplace the physical mirrors in the correct location with respect to each other.
93
Figure 3-11. Preliminary solid model of a Imager 4. The mirror has a spherical surfacewith a radius of curvature that best approximates its true aspheric shape.
94
Figure 3-12. A 3-D solid model of the opto-mechanical layout of CIRCE with the correctoptical surfaces, locations, and rotations.
95
Figure 3-13. Two practice mirrors, Imagers 2 and 3, produced by the UF Physics andAstronomy Machine Shop.
96
Figure 3-14. A typical pattern used to lightweight Fold 1, Fold 2, Collimator 1, Imager 1,and Imager 4.
97
Figure 3-15. Two different views of a lightweight solid model of Imager 4. The image tothe left shows the back surface and the image on the right shows the front.
98
Figure 3-16. The final mechanical drawing of the blank for Collimator 1. The mechanicaldrawings for all eight CIRCE mirrors are currently at the UF Physics andAstronomy Machine Shop awaiting manufacture.
99
Figure 3-17. The final mechanical drawing of Collimator 1 sent to Janos Technology. Icompleted similar drawings for all eight CIRCE mirrors
100
Figure 3-18. Examples of an L and T bracket. These were used for the smaller CIRCEoptics
101
Figure 3-19. This figure shows a redesigned bracket for Imager 4. We removed the centralback gusset and inserted two thinner gussets at the side. This added stabilitywhile decreasing weight.
102
Figure 3-20. A front and back view of the bracket for Imager 2
103
Figure 3-21. The compound bracket that supports Imager 4 and Collimator 1
104
Figure 3-22. The final mechanical drawings for the bracket for Fold Mirror 1.
105
Figure 3-23. Two of CIRCE’s brackets with their “practice” mirrors.
106
Figure 3-24. A preliminary design of the CIRCE lightweight bench. The concept for theCIRCE bench was later modified by Miguel Charcos to allow the UF Physicsand Astronomy Machine Shop to manufacture it in-house (see Chapter 8)
107
Figure 3-25. A solid model of CIRCE’s optical bench, brackets, and mirrors.
108
CHAPTER 4CRYO-MECHANICAL DESIGN OF THE PUPIL BOX MECHANISM
4.1 Introduction to Cryo-Mechanisms
In Chapter 3, I discussed the mechanical design of CIRCE’s mirrors and brackets.
Once manufactured, bolted into place, and sealed inside the CIRCE dewar, these
components become static; they do not move with respect to the optical bench. In this
chapter, I describe the design of CIRCE’s cryo-mechanisms – moving parts and machines
that operate in a cryogenic environment. In particular, I focus on my design of the pupil
box assembly. Positioned directly aft of the collimators (see §2.1) at the location of the
exit pupil, the “pupil” box contains rotating Lyot, filter, and grism wheels. These wheels
hold a variety of optical elements including [1] grisms for spectroscopy, [2] a Wollaston
prism for polarimetry, [3] narrow- and broad- band filters for imaging, and [4] masks for
engineering tests. Ultimately the goal of the pupil box assembly is to allow observers
to switch seamlessly between CIRCE modes without warming up the instrument and
manually changing elements.
The pupil box itself, sometimes called a “filter” or “filter wheel” box is a enclosure
made of 6061-T6 aluminum that houses the rotating Lyot, grism, and filter wheels, as
well as the switches and motors used to control them (see Figure 4-1 for an example
of a pupil box). The back of the box is covered by a lid, which protects the filters and
grisms inside from stray and scattered light. To allow collimated light into and out of the
box, the front and back faces have “pass-through” holes that are aligned with the optical
axis. Another larger hole on the front of the box allows access to the interior mechanisms
without removing the back cover.
Each filter, grism, or lyot wheel in the pupil box is a gear. Every gear has multiple
circular holes for grisms, filters, and masks at evenly-spaced angles and a single central
hole for an axle (see Figure 4-2). This axle runs through all of the wheels and is attached
to the front and back of the pupil box. Motors, modified to operate in the extreme
109
cold, advance the filter, lyot, and grism wheels. The wheels turn to the locations for the
specified filter, grism, or prism required, aligning with the “pass-through” holes in the
front and back of the CIRCE pupil box. This allows light from Collimator 2 to propagate
into the box, through the optical element, and out of the box, where it will continue on to
Imager 1, as shown in Figure 4-3.
One of the key ideas for CIRCE was to design and manufacture the cryo-mechanisms,
dewar, and electronics using drawings and/or parts from previous UF instruments. The
obvious advantage of this plan was that we could save time by modifying already existing
designs or assemblies instead of starting from scratch. The CIRCE pupil box benefited
from this philosophy; the final design of the assembly is a hybrid of the Flamingos 1 and
2 (Elston et al. 2003; Eikenberry et al. 2006b) pupil boxes. While the size of the wheels,
number of teeth on the gears, and other crucial dimensions changed to accommodate the
differences between the optical designs, the basic structure of the CIRCE box and internal
mechanisms is very similar to its predecessors.
4.2 Optical Elements in the Pupil Box
The first step in designing the pupil box is to determine the number and type of
optics that it will hold. To meet the needs of the GTC community, we will include the
following nine filters in CIRCE: J, H, and Ks broad-band filters, JH and HK filters for
use with the grisms, 2.16 µm Brγ and 2.06 µm HeI narrow-band filters, and 2.14 µm and
2.03 µm narrow-band Kcont filters. We will also place a Wollaston prism for polarimetry in
one of the filter wheels. While the Wollaston prisms are traditionally housed in the grism
wheel, CIRCE is also equipped for spectropolarimetry. Since this requires that light pass
first through the prism and then a grism, the Wollaston prism must be fore of the grism
wheel.
The CIRCE grism wheel will have two grisms, both of which share the grating ruling
masters custom-built for Flamingos-2. The Lyot wheel will contain several masks, which
are stiff pieces of material machined with hole patterns, which we will use to test for
110
distortion, focus, and image quality during integration and engineering runs. The Lyot
wheel will also house the pupil stop: a 55 mm diameter hole cut into a mask. This will act
as CIRCE’s cold stop (see §2.1).
4.3 Design of the Pupil Box Components
4.3.1 The Filter, Grism, and Lyot Cartridges
The first step towards completing our pupil mechanism was determining the size
of the filters, masks, and grisms in our wheels. ORA designed CIRCE with a 55 mm
diameter collimated beam at the exit pupil (see §2.1). Accounting for the extra space
necessary to mount the optics into the wheels and still have a 55 mm clear aperture, we
planned to purchase ∼ 60 mm diameter filters. However, NIR filters are not off-the-shelf
items; specially ordered these optics can cost > $10,000 each and take months to deliver.
Fortunately, Research Electro-Optics (REO), an optics manufacturer, had just completed
producing filters for a large order and had an extra set of J, H, and Ks filters with a 70
mm diameter. Since these cost less than half the price of specially ordered filters, were
immediately available, and would not negatively impact the CIRCE image quality, we
purchased them. I show a picture of one of the three broadband filters currently in-hand
at UF in Figure 4-4. Since the grism and Lyot masks must be custom designed with no
option for a serendipitous purchase, the 70 mm filters set the scale for the remaining
optical elements.
Once we the knew the size of the CIRCE filters, grisms, etc., I designed a method
to mount and secure each element into its respective wheel. I chose the solution utilized
by Flamingos-1: filter cartridges. Using the 70 mm filter diameter as a guide, I scaled up
existing Flamingos-1 designs and, using AutoCAD, created a solid model of a cartridge.
I show an example of both my preliminary and final design in Figure 4-5. The bottom of
the cartridge has a large clear aperture that allows light to pass through to the optical
element. The filter rests on a small lip that surrounds the hole. The lid of the cartridge is
similarly designed. Tightly fitting, the rim of the lid presses the filter into the cartridge.
111
A soft compressible material is positioned between the metal rims of the cartridge and
the glass optic; this prevents scratching and also helps to insure that the element does not
shift inside the cartridge.
Each optic will have its own cartridge, which will slide into large holes machined
directly into the Lyot, filter, or grism wheels. My preliminary designs of the cartridge
included a wide rim that bolted to the wheel (see Figure 4-5A). After consulting with the
IAG, I opted for the the semi-circular tabs shown in Figure 4-5B. These took up less space
but were as functional as the bulkier rims.
4.3.2 The Filter, Grism, and Lyot Wheels
Before I began the design of the filter, grism, and Lyot gears, I assessed the space
available on the optical bench for the pupil box assembly. In Figure 4-3, I show the
approximate location of the pupil box in the CIRCE optical layout. The box is very close
to light rays passing between the first and second collimators. To avoid blocking the rays,
I needed to carefully limit the size of the box and therefore the size of the wheels. With
the size of the filters, grisms, and prisms set, our only available option was to decrease the
number of holes in each wheel.
This solution is not without issues; fewer elements per wheel result in more wheels,
driving up the cost and complexity of the pupil box. Using the optical layout and the size
of the filters as a guide, I found that the best compromise is a ∼ 9.5 inch diameter wheel
with five holes. Four of the spaces hold optical elements while one hole remains empty
to allow light to pass through to other wheels. After considering the number of filters,
grisms, prisms, and masks discussed in §4.2, I decided that CIRCE would contain five
wheels: 3 “filter” wheels, 1 Lyot wheel, and 1 Grism wheel. More information about these
five wheels and the filters they will house appears in §4.4.
Using a wheel diameter of 9.375 inches, large enough for five of the cartridges
designed in §4.3.1 and less than the 9.5 inch limit calculated from the optical design, I
made a solid model of a filter wheel. I included a large central hole to accommodate a
112
filter gear hub and bearing race. I present the solid model of the filter wheel in Figure 4-6
and the 2-D drawings for manufacturing purposes in Figure 4-7.
4.3.3 Wheel Hubs and Bearing Races
In both Flamingos-1 and Flamingos-2, the filter and grism wheels slide onto a central
axle inside the pupil box. “Belleville” washers at the back of the box act like a spring
to compress all five wheels until they are almost touching (see Figure 4-8) ensuring
that they do not wobble or slide up and down the axle as they rotate. However, this
close contact presents a problem. As a motor turns one of the wheels, friction causes
the neighboring wheels to turn as well. To prevent this, small sapphire balls are placed
between the wheels as spacers. Mounted in a bearing race, circular grooves machined into
the aluminum surface of the wheels, the low friction balls are free to rotate and revolve.
When a cryogenic motor turns one wheel, the rotational energy that would normally be
imparted to the neighboring wheels is instead dissipated by the motion of the balls in the
bearing race.
I used these same concepts for the CIRCE pupil box assembly. I designed a filter
wheel hub, shown in Figure 4-9, with a bearing race machined into its front and back
surfaces. This hub bolts directly into the center of the filter wheel, as shown in Figure
4-10. The hole in the center of the hub will accommodate the 0.25 inch steel axle.
4.3.4 Filter, Grism, and Lyot Gears
Each of the five wheels in the CIRCE pupil box is a spur gear with hundreds of teeth
at its edge. These teeth mesh with teeth on one of five smaller “drive” gears attached via
an axle to a cryogenic stepper motor (see Figure 4-11). These electric motors are capable
of very fine motions called “steps”, each of which advance the axle and drive gear 3.6
degrees. Thus a full revolution of the drive wheel requires 100 steps. As the teeth of each
drive gear are meshed with the teeth of a Lyot, grism, or filter wheel, “stepping” a motor
advances one of the larger geared wheels, moving filters, grisms, and prisms into and out
of the instrument’s optical path.
113
The first step in designing the Lyot, filter, and grism wheels was to determine the
number of teeth necessary for each gear. On the advice of senior engineer Jeff Julian,
we used a gear pitch of 48. A gear pitch is the number of teeth per diameter of the gear.
Using this specification, I calculated that the 9.375 inch wheels in our pupil box would
each have 450 teeth. I then chose a 0.50 inch diameter stock precision gear with the same
pitch and 24 teeth. This produced a gear ratio of 18.75.
Using AutoCAD, I created a solid model of a filter gear which I show in Figure 4-12.
Since manufacturers often produce a hole in the wheel “blank” to hold it steady as they
create the precision teeth, I included a central bore hole in the exact location of the filter
wheel hub. I sent my solid model and completed 2-D drawings to Ascent Gears. They
manufactured the blanks and sent them back to UF. I checked the finished product,
measuring the diameter and number of teeth. I show a picture of a finished gear in Figure
4-13.
4.3.5 Pupil Box and Motor Mounts
After designing the interior components of the pupil box mechanism, I began my
work on the pupil box itself– the enclosure that would house the wheels, drive gears,
motors, and axles. In Figure 4-14 and Figure 4-15, I show my final two- and three-
dimensional design of the pupil box. The box has an arch-shaped main body with a 9.75
inch diameter, large enough to accommodate the 9.375 inch CIRCE filter wheels with
a 0.1875 inch gap on either side. The walls of the box are also 0.1875. This was thick
enough to provide stiffness for the box and thin enough to avoid adding unneccesary
weight. The arch contains three holes: one in the center of the arch for a pupil box hub
(similar in design to the wheel hubs), one left of center for light from the collimator to
pass through to the filters, grisms, and prisms, and one at the top of the box for quick
access to interior without removing the lid.
Connected to the arch-shaped main body is a rectangular protrusion designed
especially for the motor mounts. I utilized dimensions from existing Flamingos-1
114
two-dimensional drawings of the motor mounts to create solid models in AutoCAD; I
show these models in Figure 4-16. CIRCE will use modified Portescap stepper motors,
the same motors used for Flamingos-1. The position of the motors and motor mounts is
entirely dependent on the size of the gears; each 0.50 drive gear must be near enough to
its corresponding filter, grism, or Lyot wheels to ensure the teeth of the gears mesh. I
carefully positioned the holes in the body of the pupil box for the motor mounts to ensure
the drive gears were 0.25 inches from the filter, grism, and lyot gears.
Since the rectangular protrusion on the pupil box was not large enough for all the
motor mounts to bolt to the front of the box without overlapping each other, I planned
to place two of the mounts on the back lid. I mirrored the motor mounting holes from
the front surface onto the lid to ensure the two attached drive gears would also be in the
correct location with respect to the larger gears. Figure 4-17 shows the filter box with the
front three motor mounts attached.
Finally, I created a 0.25 inch thick rectangular base for the pupil box that will
bolt onto a mounting plate on the optical bench. The base contained holes for 13 bolts
and 2 pins. The bolts ensured that the box would not move with respect to the optical
bench once it was mounted in place. The pins, as with the optics, guaranteed precision
alignment. This was important, as incorrect placement of optical elements inside the
box could result in obscuration of the beam, reflections from the filters, and poor image
quality.
4.4 Integration of the Pupil Box Mechanism Designs
After completing the filter, grism, and Lyot wheels, I identified which filters, grisms,
prisms, and masks would occupy which wheels. I also determined the order of the wheels
within the box. I included several blank spaces for four additional narrow-band filters
requested by scientists within the GTC community.
The first wheel contains the narrow-band filters, 2.16 µm Brγ and 2.06 µm HeI
narrow-band filters, and 2.14 µm and 2.03 µm narrow-band Kcont filters. This wheel is
115
mounted “backwards” so that filter cartridges are easily accessible through the large hole
in the front of the pupil box. I placed narrow-band filters, the most likely to be exchanged
over CIRCE’s lifetime, in this wheel since it is the most accessible.
The second wheel contains the J, H, and Ks broad-band filters. It also contains the
“dark” filter, which observers can use for taking dark exposures. It can also be placed
in position at the end of the night to “home” the instrument. Since this wheel is also
inaccessible without opening the pupil box, I filled it with filters that should not need
replacing or switching.
The third wheel is the Lyot wheel. It is inaccessible without opening the pupil box.
This wheel contains a cold stop, a Hartmann mask used for testing the instruments focus,
the Wollaston prism for polarimetry, and a blank space for a narrow-band filter. The
fourth wheel contains the JH and HK filters for spectroscopy. It has two free spaces to
accept two of the additional narrow-band filters proposed by GTC collaborators.
The fifth and last wheel in the box is the grism wheel; grisms are thick (3 - 5 inches)
and are traditionally placed at the back of the pupil box where they will not interfere with
the other optics as the wheels turn. Since I wanted the ability to replace filter cartridges
in the wheel in front of the grisms, for example to add a narrow-band filter, without
taking apart the pupil box assembly, for example to add a narrow-band filter, I modified
the wheel; I removed the space reserved for the “blank” and replaced it with a hole large
enough to pass a filter cartridge through. I show the new grism wheel in Figure 4-18.
In Figure 4-19, I present the completed CIRCE pupil box assembly with motor
mounts, filter, grism, and Lyot wheels, cartridges, and the pupil box.
116
Figure 4-1. Example of a filter wheel assembly
117
Figure 4-2. A rough sketch of a filter wheel with 5 holes for filters, grisms, prisms, ormasks, a filter wheel hub, and a hole for an axle.
118
Figure 4-3. The CIRCE optical layout with the sketch of a pupil box included. The top ofthe pupil box is very close to rays traveling between the first and secondcollimator. This space constraints were an important consideration whendesigning the wheels and box.
119
Figure 4-4. The CIRCE H-band filter as delivered from REO.
120
Figure 4-5. These are our preliminary and final designs of a filter cartridge. A) is slightlyheavier with a bulky rim at the top of the cartridge B) is streamlined andlighter with tabs for bolts.
121
Figure 4-6. This is a solid model of a CIRCE pupil wheel. Note the five large spaces foran optical element and a central space for a filter wheel hub.
122
Figure 4-7. This is an example of a mechanical drawing for a CIRCE pupil wheel. Thedetails in the pupil wheel will be manufactured in the Physics and AstronomyMachine Shop at UF.
123
Figure 4-8. This drawing shows inside of a hypothetical pupil wheel box. The pupil wheelsare mounted on an axle and separated by sapphire ball bearings. A Bellevillewasher, a compression washer that acts like a spring, holds the wheels intension.
124
Figure 4-9. A solid model of a pupil wheel hub. This hub fits into the central hole of apupil wheel. It has two main features - a tightly toleranced hole for the pupilbox axle and a bearing race. Also pictured are the sapphire ball bearings.
125
Figure 4-10. Solid model of a filter wheel with its wheel hub.
126
Figure 4-11. A simplistic model of the gears inside a pupil box. The smaller “drive” gearis affixed to an axle protruding from a cryogenic stepper motor. As the motor“steps”, the drive gear revolves, turning the larger filter, grism, or Lyot gear.
127
Figure 4-12. A solid model of a filter, grism or Lyot gear blank. Note the central bore holecreated for gear manufacturing purposes
128
Figure 4-13. A finished gear blank created by Ascent Gears. A.) is a zoomed image of theprecision teet while B.) shown the image of an entire wheel All five blanks arecurrently in-hand at the University of Florida.
129
Figure 4-14. Two-dimensional design of the CIRCE pupil box. Notice the three largeholes, one for the pupil box hub, another for light from the collimator toenter the enclosure, and a third for access into the interior of the box. Thefive smaller holes on the right side of the box are for the motor mounts.
130
Figure 4-15. Solid model of the CIRCE pupil box.
131
Figure 4-16. Solid model of a motor mount. The platform supports a Portescap motor.The bottom pedestal bolts onto the face of the pupil box. An axle attachedto the motor runs through the hollow pedestal and into the hole in the pupilbox.
132
Figure 4-17. A solid model of the pupil box with the motor mounts in place. Three of themounts attach to the front side and two attach to the lid. This was necessaryto ensure the mounts did not overlap or touch each other.
133
Figure 4-18. Solid model of the revised grism wheel. I enlarged the “blank” hole in thegrism wheel to accommodate a large cartridge. This will allow the CIRCEteam to change filters in wheel four, the wheel immediately in front of thegrism wheel, without disassembling the entire pupil box.
134
Figure 4-19. A solid model of the integrated pupil mechanism, with filter, lyot, and grismwheels, wheel hubs, axles, ball bearings, motor mounts, and pupil box
135
CHAPTER 5THE SEARCH FOR MASSIVE STARS AROUND 1806-20
5.1 Introduction
In the previous three chapters, I discussed my role in the design of the Canarias
InfraRed Camera Experiment (CIRCE), a NIR camera for the Gran Telecopio Canarias.
While, the development of cutting-edge astronomical instrumentation is a crucial step
in the scientific process, cameras like CIRCE are fundamentally driven by research. The
ultimate question astronomers ask when presented with a new tool like CIRCE is: what
science can I pursue with this instrument?
In Chapter 1, I outlined several projects possible with NIR instruments like CIRCE.
In particular, I focused on a survey to search for massive stars around the environments
of Soft Gamma Repeaters (SGRs). In this chapter, I discuss the observations, analysis,
and photometric techniques used in this study, particularly as they apply to Cl 1806-20.
In Chapter 6, I present results on my imaging of the remaining SGR environments. This
study will lead to a better understanding of SGRs and the techniques necessary to probe
their environments.
Until the recent discovery of magnetar CXO J164710.2-455216 in Westerlund 1
(Muno et al. 2006), SGR 1806-20 was the sole magnetar clearly linked to the surrounding
cluster of massive stars, namely Cl 1806-20 (Corbel & Eikenberry 2004; McClure-Griffiths
& Gaensler 2005; Bibby et al. 2008). Discovered by Fuchs et al. (1999) and further
characterized by numerous authors (LaVine et al. 2003; Eikenberry et al. 2004a; Figer
et al. 2005; Bibby et al. 2008), Cl 1806-20 is home to a variety of interesting and rare
objects, including a Luminous Blue Variable (LBV), multiple Wolf Rayets (WRs), and
several OB supergiants. However, while previous studies have explored individual stars in
the region surrounding SGR 1806-20, no concerted effort to fully characterize the cluster’s
massive stellar population has materialized. Without these data it is difficult to place
136
absolute limits on the mass and age of the cluster, both necessary to constrain the nature
of the SGR progenitor and star formation history of the cluster.
Utilizing near-infrared (near-IR) narrow-band imaging, I search for emission lines
indicative of stellar winds in massive stars (Figer et al. 1997; Hanson et al. 1996) in
an ∼ 9′ × 9′ region surrounding SGR 1806-20. Specifically, I will use a Brγ 2.16µm
narrow-band filter. In rare, evolved massive stars like WRs and LBVs, Brγ 2.16µm
emission lines produced by large stellar winds can exceed ∼ - 40A (LaVine et al. 2003;
Figer et al. 2005). Furthermore, since the Brγ emission is located in the NIR as opposed
to the optical, it is able to penetrate the dusty, extinguished regions surrounding SGR
1806-20 better than Hα emission typically used to search for massive clusters. Finally, this
near-IR narrow-band study also uncover OBI cluster stars with Brγ absorption, with more
moderate Brγ absorption ∼ EW = 5A (Hanson et al. 1996).
5.2 Detecting Emission and Absorption Lines with Narrow-band Photometry
While SGRs offer a small, discrete, and well-defined source sample, seemingly ideal to
test theories regarding their formation and evolution, the environments surrounding these
objects present a number of challenges. All four SGRs are found in areas with high stellar
densities and large or patchy extinction. In particular, SGR 1806-20 is located in a heavily
crowded region of the Galactic Plane with AV ∼ 29 mag (Corbel & Eikenberry 2004). In
an effort to circumvent these issues, I undertook a near-IR narrow-band imaging survey.
I concentrated my efforts in the near-infrared to counter the effects of extinction. Since
AK ' 0.112AV (Rieke et al. 1985), observations in this bandpass reduce the impact of
reddening inherent in optical studies, greatly increasing the number of observable sources.
Using narrow-band photometry to probe for Brγ emission or absorption indicative
of massive stars has several advantages. First, detection of emission lines would be less
ambiguous than spectral typing based on J, H, K photometry alone. Often, when using
the later method, it is difficult to tell a WR star with AV ∼ 30 mag from an M star with
AV ∼ 15 mag; narrow-band imaging has the potential to break this degeneracy, since only
137
the WR star will have emission lines. Furthermore, as mentioned in §5.1, the Brγ lines
associated with massive stellar winds have large EWs, offering an improved chance to
detect emission line flux above the continuum. However, some questions remain about the
efficacy of this method. What level of effect are we looking for? How might the effects of
reddening confuse the data?
Using two 1% filters, which transmit 1% of the total flux of a broad-band filter,
my observations sample two discrete portions of a stellar spectrum. One filter, centered
at 2.16µm, detects both the star’s continuum flux and any flux associated with a Brγ
emission line. The second, (hereafter, Kcont) centered at 2.27µm detects only stellar
continuum. Once reduction and photometry is complete, the fluxes are subtracted. If
there is an excess in the 2.16µm filter, I have detected an emission line in the selected
target. Further, the scale of the excess is proportional to the EW of the proposed line.
In an idealized case, where the continuum flux detected by the two narrow-band
filters is equal, the EW of a line is defined as:
W =Fcont − Fline
Fcont
dλ (5–1)
where W is the equivalent width of the line, Fcont is the flux from the continuum,
Fline is the flux in the line, and dλ is the band-pass. I define a value:
α =Fline
Fcont
(5–2)
and rewrite the EW as:
W = (1− α)dλ (5–3)
Then, considering m2.16, the magnitude in the Brγ filter and m2.27, the magnitude
in the Kcont filter, and assuming an idealized situation where the zeropoint fluxes and
magnitudes are equal in the two narrow-bands:
138
mexcess = m2.16 −m2.27 = −2.5 logFline
Fcont
(5–4)
A positive mexcess indicates an absorption in the Brγ or excess in the Kcont flux, while
a negative mexcess indicates an excess in the Brγ flux. Then, substituting Equation 5–2
into Equation 5–4 and solving for α yields:
α = 10−mexcess
2.5 (5–5)
Noting that a 1% filter will cover ∼ 220A in the Ks-band, I estimate that dλ = 220
A. To place an upper limit on the value of mexcess, I set W equal to the largest Brγ
equivalent width measured in the cluster to date, thus W = - 40A (Eikenberry et al.
2004a), and solve Equation 5–3 for α. Then using this value of α in Equation 5–5, I
find mexcess = −0.18 mag. This immediately sets the limit of my maximum acceptable
photometric error. To successfully identify an excess caused by an emission line with an
EW = - 40A at a 3σ level, we must perform <4% photometry. Identifying lower EW lines
(10 - 20A) requires 2% photometry.
I note that the location of my Kcont band at 2.27µm is problematic. As wavelength
increases, the ability to penetrate dust also increases. Unfortunately, at the reddening
of the cluster, the level of Brγ excess expected from massive stars is the same order of
magnitude, but in the opposite direction, as the excess caused by increased penetration of
dust between the 2.16µm and 2.27µm bands. Thus plotting my Brγ −Kcont values versus
Ks magnitude would yield confusing results; interesting emission line cluster stars would
display an excess close to zero, as the increased flux in the 2.27µm band would cancel
the emission in the 2.16µm Brγ band. However, foreground stars with a bright Ks-band
magnitude and no Brγ emission would also show no excess.
To solve this problem, I also measured the J − Ks color of each source in my study.
Since most massive stars have intrinsic J − Ks ∼ 0, their apparent J − Ks colors are
indicative of the reddening along the line-of-sight. Thus massive stars in Cl 1806-20 will
139
cluster around a particular value of J −Ks (see §5.4.4), while foreground stars would have
a significantly bluer color. Further, cluster stars without emission lines will also have a
characteristic value of Brγ − Kcont solely indicative of the increased dust penetration of
the Kcont band. Stars with large emission or absorption lines should appear grouped at the
same J −Ks value, but at an offset Brγ −Kcont.
This technique attacks a large, crowded field globally; however I am interested in
pinpointing particular stars for follow-up spectroscopy. Therefore, the results should
not be statistical – I must show that this global method can pick out distinct objects.
Cl 1806-20 provides us with the ideal opportunity to test this approach. Of all the
SGR environments, only Cl 1806-20 has extensive spectra, photometry, distance, and
reddening measurements available. By comparing my results to the known emission-line
star population, I can evaluate whether a narrow-band survey to search for Brγ absorption
and emission would indeed select these objects as follow-up candidates.
5.3 Observations and Data Reduction
On 2005 August 26-27, I used the Wide Field Infrared Camera (Wilson et al. 2003,
WIRC) on the Palomar 200” telescope to obtain J , Ks, 2.16µm Brγ, and 2.27µm Kcont
images of an 8.7′ × 8.7′ region around SGR 1806-20. Applying standard techniques
for near-infrared (NIR) observing, I used a random 9-point dither pattern with < 30′′
separation between each image to obtain the Ks, Kcont, and Brγ data. I took 3 images at
each position, resulting in 27 images. The total exposure times were 13.5 minutes in Brγ
and Kcont and ∼ 90 seconds in Ks. Similarly, I obtained 3 exposures at 5 random dither
positions with < 30′′ separation between images, resulting in a 48 second exposure in the
J-band.
I reduced the data using FATBOY, a PYTHON based data pipeline developed at
the University of Florida (Warner et al.in preparation). FATBOY performed standard
data reduction tasks, including dark and sky subtracting, flat fielding, and dithered image
combining, to produce final reduced science frames. I then performed astrometry on these
140
images using KOORDS in the KARMA software package. Using 25 stars, I tied each of
the four frames to corresponding K-band 2MASS image of the region. This produced an
astrometric match between my four science frames. The errors in the astrometry between
the 2MASS frame and each of the four bands were relatively similar, yielding error circles
with radii ∼ 0.3 ± 0.2′′. However, the astrometric match between my science frames was
better; between the Ks-band and each of the two narrow-band frames, the radius of the
error circle was reduced to 0.13 ± 0.06′′; between the broad-band J and Ks images the
error circle had a radius ∼ 0.10 ± 0.07′′.
5.4 Analysis
5.4.1 Photometry
Once I completed my data reduction, I performed PSF photometry with DAOPHOT
II and ALLSTAR (Stetson 1987, 1992). In summary, I created a model PSF using a
sample of stars isotropically distributed across each science frame. This model PSF was
assigned a magnitude, which acted as an instrumental zeropoint. Each star in the frame
was then fit with a model PSF scaled to the star’s brightest pixel. DAOPHOT then
iteratively adjusted this model to minimize error between the stellar PSF and the model
PSF and assigned a magnitude proportional to the scaling factor. Finally DAOPHOT
(and ALLSTAR) output the magnitude and error for each source.
Given the complexity of this technique and my moderate departures from the normal
procedure, I now describe my method in more detail. First, I identified stars in each of
my science images using the DAOPHOT FIND routine. I found 15000, 14000, 24392,
and 22679 stars in the Brγ, Kcont, Ks-band, and J-band, respectively, using similar
detection thresholds for all frames. Upon identifying the majority of stars in each image, I
proceeded to create a model PSF.
Ideally, the user would specify the number and limiting magnitude of stars that
DAOPHOT should use for the model PSF. DAOPHOT would then choose objects that fit
this criteria and use these “PSF stars” to create the model. During this process, the user
141
would monitor the rudimentary shape and isolation of the individual stars and discard
those that appeared problematic. These would not be used in further analyses.
However, I noted two complications that necessitated a more careful approach for
constructing a model PSF. First, the density of my fields required a concerted effort
to choose truly isolated PSF stars. In my crowded frames the potential for blending
was much higher than in less-populated fields. Furthermore, these blended objects were
difficult to detect with the limited visual inspection tools offered by DAOPHOT. Secondly,
as described in §5.2, the scale of the effect I am searching for is small, ≤ 0.2 mag, placing
strict upper limits on the acceptable photometric error. Since the quality of my PSF
photometry is heavily dependent on how well the model PSF fit the stars in the frame,
minimizing errors in the model will minimize the overall error. Therefore, I wanted to
build a PSF model using a large number of stars distributed throughout the field to
provide the best results.
With this in mind, I used DAOPHOT to identify 200 potential PSF stars in each
band. I then checked the FWHM, isolation, and roundness of each source by eye using
the standard IRAF procedure IMEXAMINE. I discarded any object that appeared
distorted or had a bright neighboring star that might confuse the DAOPHOT PSF fitting
procedure. This process generally reduced my PSF star count by 25-50%, leaving us with
100+ stars per frame, which I ensured were more or less evenly distributed across the
field.
I then created model PSFs using the DAOPHOT technique described by Stetson
(1992). First, I constructed a prototypical PSF with the aforementioned pre-selected
PSF stars and the DAOPHOT routine, “PSF”. Since stars in close proximity to the
PSF stars distorted DAOPHOT’s calculation of the model PSF, I needed to temporarily
remove or “clean” these “neighbor” stars from the frame. Using the SUBSTAR routine in
DAOPHOT, I shifted and scaled the working model of the PSF to match the position and
142
magnitude of each “neighbor” star. I then subtracted the appropriately scaled PSF from
each neighbor star.
At first, the neighboring stars did not disappear completely. This was due to the
discrepancy between the model PSF and the true PSF. Therefore, I created a new model
PSF using DAOPHOT and the more-isolated PSF stars and repeated the subtraction
process. I iterated this procedure until visual inspection showed that DAOPHOT had
removed any substantial stray signal near the PSF stars. With isolated stars in-hand, I
could then experiment to find the class of PSF (Gaussian, Moffet, Lorenz, etc.) that best
described my data.
Given my careful selection process, the class of model that best fit the PSF stars
would most likely be the optimal solution for the entire frame. The ideal model, when
scaled and subtracted from each PSF star, would leave minimal amounts of flux. However,
while a qualitative evaluation of this flux residual is possible by visually examining the
PSF subtracted images, I desired a more quantitative analysis to discriminate between the
PSF models.
To this end, I performed basic aperture photometry on each PSF star across two
frames. In both frames, the neighboring stars were “cleaned” using the process described
above. However, in one frame, the PSF star was also subtracted leaving only a residual.
I divided the residual flux by the total flux for each star and multiplied this number
by 100. If the PSF model was a poor representation of the actual star, the subtraction
was minimal and most of the stellar flux would be left, resulting in a large residual. The
remaining flux would be ∼ 100% of the total flux. If the model fit well, the subtraction
would be excellent, leaving few counts behind. The residual would be small and the
resulting percentage would be approximately zero. I calculated the value for each PSF star
and found the median. This gave us an excellent measure of how well my PSF model fit
all the PSF stars in the frame. I repeated this entire procedure on all four bands. Overall,
the percentage of flux remaining after the PSF subtraction of the best model is ≤ 1%.
143
Using the best fit PSF models for each frame, constructed using a quadratically
varying Gaussian model, we ran ALLSTAR to complete PSF photometry on all stars in
the J-band, Ks-band, Kcont, and Brγ images. Then, using the standard error Ioutput by
ALLSTAR for each star, I calculated the median error in each band. I found errors of
0.028, 0.044, 0.027, and 0.024 mag in the J-, Ks-, Kcont-, and Brγ- bands respectively. I
discuss the errors associated with this technique in greater detail in §5.4.5.
Finally, I calibrated the J and Ks magnitudes for my sources using 2-MASS
photometry. I chose 10 isolated stars with a range of magnitudes from the 2-MASS image
of my field and matched them with their counterparts in my science frames. Starting
with the Ks-band, I calculated the difference between my ALLSTAR magnitude and the
2-MASS magnitude for each star, found the median value of the offset, and applied it to
the remainder of my Ks-band data. I repeated this process for my J-band photometry.
5.4.2 Star Matching
The next step was to match the stars in the field across all four bands in order to
derive the colors and search for Brγ excess or absorption. To accomplish this I used
TOPCAT, the Tool for OPeration on Catalogues and Tables. TOPCAT is a powerful
software tool that completes a variety of tasks on tabular data; most notably it is capable
of matching several tables using user supplied criteria.
Using TOPCAT and a 0.3′′ radius, I matched Kcont (RA, Dec) to Brγ (RA, Dec)
and J-band (RA, Dec) to Ks-band (RA, Dec) for all detected stars. I calculated the 0.3′′
radius from my astrometric errors, using the error radius plus 3σ. I then matched the
two tables to each other, using Brγ and Ks-band coordinates and the same 0.3′′ radius.
Optimally, I would have matched all four bands simultaneously, however TOPCAT was
unable to perform this match directly given the size of the data sets.
I found ∼ 9000 J-band counterparts to Ks-band sources, ∼ 12500 Kcont counterparts
to Brγ sources and ∼ 7000 sources with matches across all four bands. I discuss these
statistics further in §5.4.3.
144
To test the validity of my source matching, I examined my Brγ and Kcont table. I
found 100 sources in my Brγ science image and visually identified their counterparts in
the Kcont image. I then removed the correct Kcont counterpart from the Kcont photometry,
loading this new data set, and ran my TOPCAT match algorithm. I examined the output
for false matches – instances where TOPCAT assigned a false counterpart to one of the
test Brγ sources that purposefully lacked a true counterpart. I found no incidents of
false matches. Further, I identified 100 sources in the center of my survey area, located
them by eye in each of the four frames, recorded the matches and compared my results to
TOPCAT. I found 100% agreement between the four-band match created by TOPCAT
and my visual match.
5.4.3 Limiting Magnitudes and Match Drop-Outs
Using my 2MASS calibrated J and Ks photometric catalogs and the four-band
match catalog created by TOPCAT, I determined the limiting magnitudes of my survey.
Utilizing the single band J and Ks catalogs, I plotted Figure 5-1, histograms of both
Ks-band magnitude and J-band magnitude versus number of stars at each magnitude. I
found that my plots peaked at Ks ∼ 16 mag and J ∼ 18 mag. As the previously identified
OB and WR stars in the cluster fall between ∼ 9 - 13 mag in Ks-band and ∼ 16 - 18 mag
in J-band (Eikenberry et al. 2004a; Figer et al. 2005), I confirmed that my photometric
limits were deep enough to observe the known massive cluster stars. Furthermore, given
the broad range of spectral types in the established population of massive stars and
my clear ability to detect them, I expect that my data would also include previously
undiscovered massive cluster stars at the distance of Cl 1806-20, should they exist.
I then addressed the issue of “drop-outs” between my matches. Drop-outs are
defined as sources that appear in one band, but are absent from one or more of the
remaining bands. I found a significant number of drop-outs each time I matched catalogs.
Extensive analyses (detailed below) show that the majority of drop-outs fall into one of
two main catagories: stars at the field edges missing from one or more images due to small
145
variations in the field-of-view between frames or very faint stars at my detection limits. A
smaller percentage of drop-outs were objects in the most crowded regions of the field.
I first examined the Brγ and Kcont match and noted a significant number of
drop-outs; the 2-band catalog contained 12689 objects, 1300 less than the 13989 sources
in the Kcont, and ∼ 2500 less than the 15233 sources in the Brγ. Discounting the sources
lost due to non-overlapping regions at the edge of my frames, I was left with 300 sources
in Kcont with no Brγ counterpart. I plotted a histogram of the number of sources versus
instrumental Kcont magnitude; first for the complete Kcont catalog and then for the 300
Kcont drop-outs. The plots showed that the Kcont drop-outs were systematically fainter
than the limiting magnitude, Kcont ∼ 20.3 mag. If these sources, on the threshold of my
detection limit in Kcont, were a few counts fainter in Brγ this would account for their
detection in one band and not the other. This is plausible; recall that a 2.27µm filter will
detect more flux than a 2.16µm filter in stars with no absorption given the increased dust
penetration by longer wavelengths (§5.2).
I then examined the remaining ∼ 1500 Brγ sources with no Kcont counterpart. I
created a histogram of the number of sources versus Brγ magnitude for both the entire
Brγ catalog and the Brγ drop-outs. I found a limiting magnitude of Brγ ∼ 20.3 mag and
noted that the drop-outs were on the faint end of the distribution. Again, if these sources
were only marginally brighter in Brγ than in Kcont, I would detect them in one filter and
not the other.
Using TOPCAT I matched my Brγ drop-outs to my J and Ks broad-band match.
I repeated this for my complete Brγ catalog and plotted histograms of number of stars
versus J−Ks color for both catalogs. I found that my Brγ drop-outs were bluer (J−Ks ∼1 - 2 mag) than the bulk of my J −Ks distribution. Thus, faint, slightly blue foreground
stars most likely account for the population of dim Brγ sources lacking Kcont counterparts.
None of the narrow-band drop-outs affect the goal of this survey. Luminous early-type
stars in Cl 1806-20 have apparent Ks-band magnitudes between 9 - 13 mag. Later in this
146
section, I show that this Ks magnitude corresponds to Brγ and Kcont between 14 - 18
mag; my drop-outs are much fainter than this. Further, as discussed in the paragraph
above, the J −Ks color of the Brγ drop-outs ∼ 1 - 2 mag; much bluer than the J −Ks ∼5 mag attributed to the cluster (§5.4.4).
I also observed a large number of drop-outs between the J- and Ks-bands. Out of
24392 sources in Ks and 11812 sources in J , 9300 sources were included in the matched
Ks and J catalog. The large number (> 10000) of Ks-band sources with no J-band
counterpart was unsurprising; since longer wavelengths are more efficient at penetrating
dust I expected many of the Ks stars to be highly obscured and thus undetected in the
J-band. This is visually noticeable in the final images themselves. However, once source
of concern was the large number of J-band sources with no Ks- band matches. While I
expect most sources detected in J-band to have a counterpart in Ks-band I found that
>2000 J-band sources dropped-out of the Ks-band match. Discounting the 800 sources
lost due to non-overlapping frames, I am left with 1200 dropouts.
Two factors account for the loss of J-band sources in the J to Ks match: a blue
foreground population and confusion and crowding in the KS-band image. After exploring
each of these issues, I concluded that these drop-outs in my catalog caused by these
factors would not affect my ability to detect massive cluster stars. I detail my drop-out
analysis below.
To account for the 1200 drop-outs, I chose 100 random sources from this population,
found their positions on the Ks-band image, and searched for their missing counterparts.
In ∼ 80% of the cases, a very faint object, undetected by the DAOPHOT FIND routine,
did exist in the Ks-band image. I observed that the peak counts attributed to these
sources were of the same order of magnitude as the calculated noise in the background.
Therefore, given my inputs, the DAOPHOT FIND routine did not recognize these faint
objects as sources and excluded them from the KS catalog. Extrapolating from the results
147
of my random selection of 100 stars, I estimate that 80% of my 1200 J-band drop-outs, or
∼ 1000 stars are members of this population.
I then examined the same 100 random stars in the J-band image. The stars, while
faint in the J-band, were significantly brighter than their “undetected” Ks counterparts
described above and therefore easily identified by DAOPHOT. I plotted a histogram of the
1200 J-band dropouts missing from the Ks-band catalog. I found that 1000 sources were
at the faintest end of the histogram with J > 17.5 mag.
While I could alter my input into DAOPHOT in an attempt to detect the faint
Ks-band counterparts to these dim J-band stars, this would also result in a larger number
of spurious detections. Furthermore, given that the massive cluster stars I was interested
in finding are luminous in the Ks-band (between 9 - 13 mag), while these stars are very
dim in the Ks band, the likelihood of this tactic yielding useful results was minimal. Thus
I did not attempt to include these sources in my catalog.
The remaining ∼ 20% of my J-band drop-outs are most likely due to crowding in
the Ks-band. Of the 100 random sources I inspected, ∼ 20 stars were obviously visible in
both images. However in the Ks-band image, source confusion and crowding was evident;
often a brighter star or a large grouping of stars that did not appear on the J-band
image overlapping its Ks-band counterpart. In cases where one star almost completely
overlapped another, it was unlikely that the FIND routine would recognize each star as
a separate object. Correcting this problem is not trivial; even upon visual inspection, it
was challenging to find the correct counterpart. Thus, in some cases, in some of the most
crowded areas of my Ks-band image my observations reach the confusion limit.
Finally I examined my 4-band match. As mentioned in §5.4.2, I created this catalog
by matching the Brγ to Kcont table, with ∼ 12000 sources, to the J- to Ks-band table,
with ∼ 9000 sources. A total of 7010 sources are matched across the four bands. Since one
might expect all 9000 sources with both J- and Ks- band counterparts to appear in the
narrow-band catalog I again note a significant drop-out rate (∼ 2000 stars).
148
To investigate this dropout rate, I examined the sensitivity limits of the broad-band
Ks versus the narrow-band data. Using TOPCAT, I performed an astrometric match
between the Brγ, Kcont, and Ks-band catalogs using a procedure similar to that described
in §5.4.2 . I then created a histogram showing the number of sources per Ks-band
magnitude bin versus Ks-band magnitude for the new 3-band match. I overplotted
this histogram with the one created for the Ks-band catalog. Figure 5-2 shows that a
large number of Ks-band sources with Ks magnitudes fainter than > 16 mag had no
narrow-band counterparts in my data and thus were not included in the 3-band (and
therefore 4-band) match. Thus, the lack of narrow-band counterparts to faint Ks stars is
due to the sensitivity of my narrow-band observations; the narrow-band images have a
brighter limiting magnitude for my integration times.
As mentioned in §5.4.1, I was unable to directly calibrate the Brγ and Kcont
magnitudes from published catalogs. Instead I created a histogram of the instrumental
magnitudes versus the total number of stars in each magnitude bin for both narrow-band
catalogs. I then overplotted this histogram on top of the histograms presented in Figure
5-2; the result is Figure 5-3.
The shapes of the narrow-band histograms resemble the shape of the histogram of
Ks-band sources with narrow-band counterparts. By matching narrow-band stars to
their Ks counterparts, I have effectively photometrically calibrated each source. This
is reflected in the histograms; when comparing the shape and peak of the narrow-band
histograms to the 3-band match histogram, it is evident that the Brγ and Kcont data are
offset roughly ∼ five magnitudes from the Ks-band data. Further, I note a significant
drop in the narrow-band histogram at instrumental magnitudes of ∼ 21 mag. In the Ks
calibrated scale this corresponds to Ks ∼ 16 mag.
Using this information, I conclude that the remaining source drop-outs from the J-
to Ks- band catalog to the final 4-band catalog are caused by a drop in the sensitivity of
my narrow-band data due to insufficient integration time; my narrow-band observations
149
were too short to reach the same limiting magnitude as my broad-band imaging. However,
given the expected brightness of potential massive cluster stars, (Ks < 16) as discussed in
the beginning of this section, this magnitude limit for my narrow-band imaging should not
adversely affect my results.
5.4.4 Field Selection
While my data provide information about stars across the entire field, I am primarily
interested in potential members of Cl 1806-20. Thus, from this point on, I only consider
a subset of my field – in particular a 6 ′ × 6′ subregion centered on the SGR. I chose the
size of this region after reviewing the published radii for Galactic massive star clusters
whose stellar contents, ages, and (potentially) masses are similar to Cl 1806-20 (Figer
et al. 1999, see Table 5). In particular, I sought an estimate that would not exclude less
compact clusters or ejected massive stars. Such a cluster is represented by Westerlund 1, a
potential Galactic Super Star Cluster (SSC). While the bulk of Westerlund 1 lies within a
1.2 pc core, the cluster is elongated, with several cluster members located well outside this
radius (Clark et al. 2005); in fact potential WR cluster members are located as far as ∼ 8
pc away (Crowther et al. 2006).
While I recognize that Westerlund 1 is not a direct analog to Cl 1806-20, I use
its size to place an upper limit on the search area necessary to locate members of a
massive stellar cluster. At 15 kpc, the distance to Cl 1806-20 reported by Corbel &
Eikenberry (2004) and supported by McClure-Griffiths & Gaensler (2005), a 16 pc × 16pc
search area corresponds to a subfield ∼ 4′ × 4′. However, if the cluster is at the ∼ 9kpc
distance suggested by Bibby et al. (2008), my field would correspond to a 10 pc × 10 pc
search area. To ensure that despite the disputed distance of Cl 1806-20 we still cover an
approximately 16pc × 16pc area, we use a larger 6′ × 6′ field.
We computed the J −Ks versus Brγ −Kcont values for all stars in my 4-band catalog
within the 6′ × 6′ area. The result is Figure 5-4. Then using the RA and Dec in my
4-band table for my stars and an RA = 18h 08m 39.32s and Dec = -20◦ 24′ 39.5′′ (Kaplan
150
et al. 2002) for the SGR, I calculated the 2-D projected distance between each field star
and SGR 1806-20 . These 2-D projected distances are represented by a color gradient,
with objects close to the SGR colored purple and objects further away in red.
Figure 5-4 offers a wealth of information about the stars in my field and the
populations to which they belong. In particular, the J − Ks color is a distance and
reddening indicator; objects on the left of the graph are bluer and therefore most likely
located in the foreground, while objects to the far right are heavily reddened. Since AK
= 0.112AV and AJ = 0.282AV (Rieke et al. 1985), using an AV = 29 ± 2 for Cl 1806-20,
(Corbel & Eikenberry 2004; Eikenberry et al. 2004a) I calculate AK = 3.25 ± 0.56 and
AJ = 8.18 ± 0.22 yielding a J −Ks = 4.93 ± 0.34 mag for massive stars with an intrinsic
J −Ks = 0.
5.4.5 Errors
Quantifying and characterizing the error in my photometry is crucial for the success
of this technique. In §5.2, I found that a 4σ detection of a Brγ emission or absorption line
with an equivalent width ∼ 20 - 40A required 2-5% photometry. As mentioned in §5.4.1,
ALLSTAR outputs both a magnitude and standard error for each source in the frame.
Using these data, I calculated the median photometric error in each band and found that
it ranged from 0.024 - 0.044 mag. However, while these values establish that I achieved
better than 5% photometry in each frame, they do not necessarily confirm that I met
the criterion set in §5.2. In actuality, I require the photometric error in the Brγ excess or
absorption, and thus the error in Brγ −Kcont, to meet the 4% limit.
To check if my photometry satisfies this standard, I initially used the Brγ and Kcont
2-band match described in §5.4.2. First, I calculated the median error in Brγ−Kcont to be
0.032 mag. This corresponded to ∼ 3% photometry, confirming that, on a global level, my
errors were small enough to allow for 3σ detections of modest Brγ emission and absorption
lines.
151
I then repeated this calculation using my 6 ′ × 6′ 4-band catalog and found that the
median Brγ − Kcont error dropped to 0.020 mag. Two factors account for this decrease.
As noted in §5.4.3, dimmer sources in my field often lacked matches across all four
bands and were therefore not included in the 4-band catalog. Since dimmer sources have
systematically larger Poisson errors, their absence from the catalog improved my overall
data quality in the area of interest. Further, by limiting the catalog to the central region
of the frame, where minor distortion effects inherent in the detector were at a minimum,
I effectively eliminated errors associated with poor model PSF fits. Thus, the photometry
over the most interesting part of the field exceeds my 5% criterion.
Having met the necessary photometric limits on a comprehensive level, I then
assessed my error on a more source specific scale. I computed the errors in the J − Ks
and Brγ − Kcont colors for every star in my 4-band catalog using the errors output by
DAOPHOT. If a particular source displays emission or absorption but a large error in
either color, I can use these errors to evaluate whether it is still a good candidate for
follow-up spectroscopy. I discuss this in more detail in the sections below.
5.4.6 Calculating Equivalent Widths
In Figure 5-4 I noticed a linear correlation between the J −Ks color and Brγ −Kcont
color of sources in the field. This was expected because, as mentioned in §5.2, longer
wavelengths penetrate dust more efficiently. Since my Kcont filter was centered at a longer
wavelength than my Brγ filter, it systematically detected more flux, even if no intrinsic
excess or absorption existed. However, the size of this effect depends on the reddening;
if the extinction towards a particular star in the field was small, the effect was small. If
the extinction was large, the effect was large – on the order of 0.1 - 0.3 mag. The slope of
the line is therefore an indicator of the varying reddenings and distances of objects in the
field-of-view.
Using TOPCAT, I fit a line with slope (m) = 0.028 and y intercept (b) = -0.137
to the data points. For any given J − Ks color, this line defines the expected value
152
of Brγ − Kcont for a star with neither Brγ absorption nor emission, thus acting as a
narrow-band “zeropoint”. This zeropoint is a combination of the dust-penetrating effect
described above and differences in transmission between the narrow-band filters. To find
the calibrated Brγ − Kcont value of an individual star in my data set, I first solved the
equation of the line using the star’s J−Ks value. This gave us the Brγ−Kcont “zeropoint”
for that particular object. I then subtracted this value from the photometrically derived
Brγ − Kcont color. The resulting magnitude difference was used with Equation 5–5 and
Equation 5–3 to estimate the EW of the Brγ excess or absorption. Finally, using the
errors in the Brγ − Kcont and J − Ks colors calculated for each source (§5.4.5) and the
linear equation described above, I computed the errors in the EW for every star in my
4-band catalog.
5.5 Comparison to Previous Results
By comparing prior spectral and photometric information (Table 5-1) to my observed
Brγ excess or absorption and J − Ks colors (Table 5-2), I am able to evaluate how well
this global narrow-band photometric technique estimates the properties of individual stars.
While Figure 5-4 contains substantial information about a number of stellar
populations in the field (§5.6), I sought a more useful visual tool to highlight the
previously known bright cluster members and identify new candidates. In Figure 5-5 I plot
stars with Ks ≤ 16, the completeness limit of my Ks catalog. As mentioned previously,
cluster stars should have apparent Ks magnitudes between 9 - 13 mag; by removing the
dimmer foreground stars, I limit contamination by foreground stars with large broad-band
photometric errors. Further, since faint Ks stars also have faint narrow-band counterparts,
I exclude sources with large narrow-band photometric error. This reduces the scatter
around the “zeropoint” Brγ − Kcont and makes it easier to distinguish stars with real
emission or absorption features from those sources with large narrow-band photometric
errors. Stars with Brγ excess or absorption, discussed in detail below and in §5.6.1, are
153
clearly labeled in Figure 5-7, a 1.3′ × 1.6′ region of my Ks-band image encompassing Cl
1806-20.
5.5.1 Wolf Rayet Stars
LaVine et al. (2003, in preparation), Eikenberry et al. (2004a), Figer et al. (2005), and
Bibby et al. (2008) present spectra for Wolf Rayet stars in Cl 1806-20. I summarize these
results in Table 1. For the remainder of the discussion, I adopt the LaVine et al. (2003, in
preparation) and Eikenberry et al. (2004a) nomenclature when an object has more than
one designation.
Using the finder charts and coordinates provided by these authors, I visually
identified each of the known WR cluster stars on my Brγ image. I then found the location
of these objects on Figure 5-5 (denoted as squares) and determined their Brγ EWs from
my narrow-band imaging.
Three of the four WR stars, B, 10, and 22 appear in my catalog. Star 3 is the
exception. The location of Star 3 in the densest, most confused region of the cluster,
prevented an accurate measure of its position or magnitude in one or more bands in my
data.
Star 22 and Star B show the potential of the technique described in this paper. Star
22 has an Brγ excess of - 0.20 ± 0.01 with a J −Ks error = 0.02. Using Equation 5–5 to
solve for α, Equation 5–3 to solve for W (see §5.2), and incorporating the errors described
in §5.4.5, I find an EW ∼ - 45 ± 1A, in agreement with the EW = - 42 ± 3A previously
reported (see Table 1). Also, I measure a J − Ks ∼ 5 mag for Star 22, consistent with
earlier measurements of this source (LaVine et al. in preparation).
I repeated the above calculations for Star B. Surprisingly, I found that Star B does
not have a Brγ excess; instead my data show a Brγ absorption of 21 ± 1A. While at first
this appears to be inconsistent with the previously reported Brγ emission, inspection of
the spectra provided by Eikenberry et al. (2004a) and LaVine et al. (in preparation) show
that this result is consistent with the earlier data. Star B is classified as a WCLd, a dust
154
emission WR star. Redward of 2.2µm, Star B exhibits a large IR excess. Since the Kcont
band used in my observations is located at 2.27µm, I expect the flux in this narrow-band
to be greater than the flux in the Brγ band. Thus, my apparent Brγ “absorption” is in
fact a documented Kcont excess. Further, my J and Ks- band magnitudes (Table 5-2) are
consistent with the J- and Ks- band magnitude of 17.78 and 10.5 respectively reported by
LaVine et al. (private communication).
Star 10 is an interesting enigma. LaVine et al. (2003, in preparation) report a Brγ
emission line ∼ - 25 ± 1 A. However, I find a small Brγ absorption (1 - 3A), which is
inconsistent with previous results. To rule out a spurious match, I visually identified
the object in each of the four images and verified that TOPCAT selected the correct
counterpart.
Noting that Star 10 was relatively isolated, I completed an independent check of the
DAOPHOT photometry by performing aperture photometry on the star in the Kcont and
Brγ bands. The value of Brγ − Kcont found with this method matched the Brγ − Kcont
color calculated from my PSF photometry, confirming the observed lack of excess.
One possible explanation for the absence of Brγ emission is that Star 10 is a dust
emission WR. Recent spectroscopy from Bibby et al. (2008) suggests that like Star B, Star
10 is a WC9d, with warm dust masking the strength of the emission lines. Also, with a
J ∼ 17.38 ± 0.02 mag and Ks ∼ 11.55 ± 0.01 mag, I calculate a J − Ks color ∼ 5.83
mag. This value is slightly redder than most stars in the cluster and may be indicative (as
with Star B) of intrinsic reddening by circumstellar dust, consistent with the Bibby et al.
(2008) result.
Another possible reason for the absence of Brγ excess in Star 10 is variability in the
emission line. However, follow-up spectroscopy over multiple epochs would be necessary to
test this hypothesis.
155
5.5.2 Luminous Blue Variable (LBV) 1806-20
First discovered by Kulkarni et al. (1995), the candidate Luminous Blue Variable
(LBV) 1806-20 was originally believed to be the infrared counterpart to SGR 1806-20.
Subsequent studies showed that it was not the magnetar’s counterpart (Hurley et al.
1999), but a member of Cl 1806-20 (Fuchs et al. 1999; Eikenberry et al. 2001; Corbel &
Eikenberry 2004).
While the mass range and potential binary nature (Eikenberry et al. 2004a; Figer
et al. 2004) of LBV 1806-20 are topics of contention in the literature, the intensity of the
star’s massive stellar winds is well established. Eikenberry et al. (2004a) report several
strong emission lines, particularly Paβ ( ∼ - 82A) and Brγ (∼ - 44A). Spectra in Figer
et al. (2004, see their Figure 3) show similarly large emission lines.
Given the reported level of Brγ emission, I expect to find a large Brγ EW for LBV
1806-20 within my data. However, using the calculations described in §5.4.6 and my
Brγ − Kcont excess, I find an EW ∼ -4 ± 2A. I investigate two potential reasons for this
discrepancy: 1) errors in my photometry or 2) intrinsic variability of the emission line.
5.5.2.1 Evaluation of PSF photometry
While global errors in my data met the 2 - 5% photometric limit necessary to detect
20 - 40A Brγ emission lines (see §5.2), the nature of this technique also allowed us to
evaluate whether this criterion was met on a source by source basis (§5.4.5). I found that
the standard error in each band for LBV 1806-20 was small (< 0.01 mag); my photometric
errors could not account for an inconsistency of ∼ 40A between my measurements of the
Brγ emission and that of Eikenberry et al. (2004a).
As a check on the DAOPHOT photometry, I performed aperture photometry on LBV
1806-20 in my Brγ and Kcont science frames using the procedure described in §5.5.1. I
also completed aperture photometry on three other, isolated spectroscopically identified
sources: Star 5 (a red giant), Star 22 (a WR), and Star C (an OBI). I then computed
the value of Brγ − Kcont for each source and compared it to the values from my PSF
156
photometry. I found that the Brγ − Kcont colors derived from aperture photometry for
Stars 2, C, 5, and LBV 1806-20 were consistent with the PSF photometry to within the
PSF derived photometric errors. Thus my PSF photometry is consistent with simple
aperture photometry. I conclude that the absence of the expected Brγ excess is not due to
errors in my data or analysis.
5.5.2.2 Spectroscopy of LBV 1806-20
Variability of wind-driven lines in massive stars is not uncommon; LBVs, in
particular, often show substantial spectroscopic variation in many emission lines
including H, HI, and FeII (Humphreys & Davidson 1994). I explore the possibility
that intrinsic variability accounts for the significantly lower than expected Brγ emission in
my narrow-band image.
J. LaVine and V. Mikles obtained spectra of LBV 1806-20 on 17 May 2004 and 2
July 2005 using the near-infrared spectrograph SpeX (Rayner et al. 2003) on the InfraRed
Telescope Facility (IRTF). For the data taken in 2004, they acquired twelve 60s images of
LBV 1806-20, for a total exposure time of 720 second. For the 2005 run, they took six 90s
images, for a total exposure of 540 seconds.
I reduced these data using standard SpeXTool procedure and produced flat-fielded,
sky-subtracted, wavelength calibrated spectra for LBV 1806-20 for each of the two nights.
Then, using Xcombspec (Cushing et al. 2004) I combined the individual images of the
LBV to produce two weighted-mean LBV spectra, one for 2004 and one for 2005. Finally,
I divided the corresponding atmospheric standard from the program star, multiplied
the LBV spectra by a 5600 K blackbody spectra (the temperature of the standard),
dereddened them using Av = 29 mag and calculated the EW of all emission lines.
My resulting K-band spectra appear in Figure 5-7. A list of identified emission lines
appear in Table 5-3. The EW of the spectral lines are consistent between the two epochs
of observation. The only exception is Brγ; however the difference in the EWs between the
two epochs is less than 3σ. More interesting are the differences between my 2004/2005
157
data and the 2001 spectrum of Eikenberry et al. (2004a) and 2003 spectrum of Figer et al.
(2004). Both of the earlier spectra show large Brγ emission lines; Eikenberry et al. (2004a)
reports an EW = ∼ - 40A. However, I find Brγ EWs of ∼ - 7.9 ± 0.4A in May 2004 and
- 9.13 ± 0.6A in July 2005. Even accounting for errors, this is a > 50σ difference. Other
lines show similar variability: e.g. while I find an emission line consistent with zero for HeI
(2.053µm) Eikenberry et al. (2004a) report EW = - 17.4 – again, a > 20σ difference.
The Brγ emission detected in my 2004/2005 spectra is significantly weaker than the
Brγ emission in spectra observed in 2001/2003 confirming variability in the Brγ line.
Furthermore, the Brγ EW derived from my 2004/2005 spectra are of the same order of
magnitude as the 4 ± 2A Brγ EW found by my narrow-band imaging survey. Obtained
∼ 15 and 2 months after my spectroscopy, the similarities in the imaging derived Brγ
EW indicate that my narrow-band photometry is measuring a spectroscopically confirmed
intrinsic variability in the Brγ emission line of LBV 1806-20.
My broad-band photometry reinforces this finding. I find Ks = 8.56 ± 0.01 and J =
13.45 ± 0.02 for LBV 1806-20, compared to K = 8.89 ± 0.06 and J = 13.93 ± 0.08 found
by Eikenberry et al. (2004). Taking into account the errors and slight difference between
the K- and Ks- band, my Ks measurements is noticeably brighter. This is in agreement
with reported variability in LBVs, which often show strong line emission at photometric
minimum and weaker lines at higher luminosities (Humphreys & Davidson 1994).
Further, given that the EW of the Brγ line may have varied by a factor of four
between 2001/2003 and 2004/2005, it does not seem unreasonable that the Brγ flux
could decrease by another factor of two between my 2005 spectrum and the narrow-band
imaging 2 months later. Taken together, these data present clear evidence for variability of
the Brγ emission over various timescales.
5.5.3 OBI stars
Eikenberry et al. (2004a) identify two OBI stars in Cl 1806-20: Stars C and D. Figer
et al. (2005) confirmed these identifications and found another OBI star, Star 4, and two
158
OBI candidates, Star 7 and Star 11. Recently Bibby et al. (2008) obtained spectroscopy
confirming all five objects as OBI stars and assigning more accurate spectral types (see
Figure 5-1). While none of these authors give exact measurements for the EW of the
Brγ absorptions, Figer et al. (2005) estimates an EW = 3 - 5A for C, D, and 4 and no
noticeable Brγ absorption in the spectra of Stars 7 and 11. Bibby et al. (2008) suggest
their spectroscopy is consistent with the Figer et al. (2005) results.
I identify all five OBI stars on Figure 5-5 (labeled with triangles). Then, I estimated
the Brγ − Kcont color of each object as described in §5.4.6. Finally, inserting the
narrow-band color and errors into Equation 5–5 and the resulting value of α into Equation
5–3, I calculated the Brγ absorption. The results are presented in Table 5-2.
For Star 4, I found a Brγ absorption line with an EW = 20 ± 5A, four times that
of Figer et al. (2005) and considerably larger than the line observed by Bibby et al.
(2008, see their Figure 1). Located in the densest portion of the cluster, it is clear upon
inspection that I have reached the confusion limit. I count several stars within my 0.3′′
radius error circle and find it difficult to visually identify individual sources. Upon
examining my subtracted DAOPHOT frame, I see that the program did not accurately fit
Star 4 and the stars surrounding it correctly with the model PSF. This is a limit of my
data and not the technique; with adequate resolution in future observations, I anticipate
that I could solve these problems. Thus due to the crowding in the field, the uncertainty
associated with my reported EW for this star is larger than the formal errors.
My values for the EWs of Stars C and D, while two times larger than the range
reported by Figer et al. (2005), support those of Eikenberry et al. (2004a), who suggested
that these objects were hypergiants: OBI stars with abnormally strong absorption lines.
Furthermore, Bibby et al. (2008, see their Figure 1) show moderate absorption in Star C,
consistent with my findings. Noting the lack of specific EW measurements for Star C and
D in all three previous works, further discussion of these source must wait for follow-up
spectroscopy.
159
In addition to the absorption detected in Stars C, D, I also found EW ∼ 5 ± 1A for
Stars 7 and 11, despite previous reports of no Brγ absorption (Figer et al. 2005). Since
both stars were isolated, I performed aperture photometry on Stars 7 and 11 similar
to that performed on Star 10. I found an EW ∼ 6A for Star 11 and an EW ∼ 2A for
Star 7. Both sources of photometry for Star 11 are consistent and confirm a small Brγ
absorption. While my aperture photometry of Star 7 predicts a slightly lower EW than
that found from my PSF photometry, errors in the aperture photometry (> 5%) exceed
this difference. Thus, the values derived from both methods are in agreement. Comparing
my measured values qualitatively to the spectra of Stars 7 and 11 presented by Bibby
et al. (2008) (who do not provide EW measurements) suggests that my detection of a
small Brγ absorption is consistent with their spectra.
5.6 New Results on Observed Stellar Populations in My Field
5.6.1 New Candidate Cluster Members
Having carefully compared my results for all cluster members identified to date in
the literature, I find that my technique can identify potentially interesting objects. Most
OBI and WR stars in the cluster show evidence for Brγ absorption or emission, marking
them as interesting spectroscopic targets. Further, I am able to make reasonable estimates
regarding the strength of the absorption or emission in these sources.
Assuming a reddening to Cl 1806-20 of AV = 29 ± 2 (Corbel & Eikenberry 2004;
Eikenberry et al. 2004a)), I calculate a color of J − Ks ∼ 4.93 ± 0.34 mag for massive
stars with no intrinsic reddening in the cluster. To search for new candidate cluster
members, I examine this region of Figure 5-5, expanding my search to J − Ks ∼ 4.5 -
5.5 mag to account for potential patchiness in the field (Corbel et al. 1997; Corbel &
Eikenberry 2004), potential variations in the reddening law within the field (see Gosling
et al. in preparation), and photometric errors in J − Ks color. Finally, this range better
encompassed the J −Ks color of the known OBI cluster population.
160
Upon careful inspection of Figure 5-5 in the J − Ks ∼ 4.5 - 5.5 mag color region,
I observed no large population of stars with a noticeable Brγ emission. If a significant
number of undiscovered emission line WRs existed, I would expect them to cluster near
Star 22, yet my results show Star 22 to be relatively isolated. This suggests that Cl
1806-20 may not contain many undiscovered extreme WR stars.
On the other hand, I did observe ∼ 10 Brγ absorbers in the region of the color-color
diagram that contained several previously known OBI cluster members. These objects,
potential OBI stars, are the most interesting candidates for follow-up spectroscopy and
may represent the bulk of previously undiscovered massive star population in this cluster.
I present the broad-band colors, Brγ absorption, and distance from the SGR in projected
distance in Table 5-4. I note that a few of these objects are far from the position of the
SGR; these could be massive stars not associated with Cl 1806-20 or OBI stars ejected
from the cluster.
5.6.2 Foreground Stars
Given a J − Ks color of ∼ 5 mag for Cl 1806-20 (see §5.4.4) I expect a significant
portion of stars in Figure 5-4 and 5-5 to belong to foreground populations. Given the
bluer color of objects on the far left of Figure 5-4, this description particularly applies to
stars in the J −Ks ∼ 0 - 2.5 mag color range.
The spread in narrow-band colors for this foreground population is rather large, with
the bulk of the stars lying between Brγ −Kcont = -0.2 - 0.0 mag. I note that the majority
of my faint Ks stars, with systematically larger photometric errors in all four bands, reside
in this portion of the graph; thus there is more scatter in the narrow-band colors in the
foreground population.
Most stars in the J − Ks = 2.5 - 4.5 mag region of Figure 5-4 also fit the definition
of foreground star; compared to my cluster, they are notably bluer. However, given the
patchy extinction in the direction of Cl 1806-20, it is also possible the some objects on
the redder end of this distribution could be cluster stars. Unfortunately with the small
161
number of known cluster members it is very difficult to statistically determine if (for
instance) an individual star in the J − Ks = 4.3 mag region of the plot is a potential
cluster member. In the end, stars with large Brγ emission or absorption and broad-band
colors similar to that of the cluster should be examined on a case-by-base basis.
Several interesting features are apparent in Figure 5-4, most notably the segregation
of foreground stars into two discrete distributions centered at J − Ks ∼ 1.5 mag and
J −Ks ∼ 3.7 mag. I will further explore these foreground stellar population along the line
of sight to the cluster in the future.
5.7 Conclusion
I have demonstrated that a narrow-band imaging survey can be a useful method of
searching for massive stars in a crowded field. Applying this technique to the environment
of SGR 1806-20 yielded several results:
1) I confirmed the existence of known massive stars in Cl 1806-20, the host cluster
of SGR 1806-20. Several of my reported EWs are in good agreement with the literature
values. Where discrepancies exist, I explored the reasons. I found that in some cases,
insufficient information in the literature prevented quantitative comparison. In other cases,
the source of the discrepancy may be a result of intrinsic variations. In one case crowding
in the field yielded a poor result. From this I concluded that better resolution is necessary
to probe the most crowded regions of the cluster.
2) My narrow-band observations of LBV 1806-20 did not match the expected Brγ
emission. I presented recent spectroscopy that confirms that the Brγ line is variable in
LBV 1806-20. This suggests that the discrepancy between my imaging-derived EW width
and that reported by Eikenberry et al. (2004a) may be a result of intrinsic variability
in the Brγ emission line. Further systematic observations are required to determine the
characteristic timescales and magnitude of the spectroscopic variability.
3) I did not detect any significant previously unknown WR population in Cl 1806-20.
However I have identified a population of candidate OBI stars that may represent the bulk
162
of the missing cluster population. I suggest that these stars should be targeted for future
spectroscopic observations.
163
Table 5-1. Summary of previous work on massive stars in Cl 1806-20. Columns give theID, Spectral Type, and Brγ EW of Wolf Rayet stars in Cl 1806-20. All data aretaken from literature sources. Values in parenthesis indicate uncertainties inthe last digit of the EWs.Numbers indicate the nomenclature and reportedspectral types of each author where discrepancies occur; when no numbers aregiven, all authors listed below are in agreement
Massive Star Spectral Type Brγ EW(A)
Star B WC9a,b,d ; WC9de -11(1)b,d
Star 10b,d ; Star 1c,e WC9b,d ; WC8c ; WC9de -25(1)b,d
Star 22b,d ; Star 2c,e WN4b,d ; WN6 c ; WN6be -43(3) b,d
Star 3c,e WN7c,e –LBV 1806-20a LBV -44.2(3)a
Star 4c,e 09.5Ie –Star 7c,e B0I - B1Ie –Star 11c,e B0Ie –Star C B1 - B3Ie –Star D OBIe –a Eikenberry et al. 2004ab LaVine et al. 2003c Figer et al. 2005d LaVine et al. in preparatione Bibby et al. 2008
164
Table 5-2. Measured magnitudes, colors, and Brγ excess or absorption for massive stars inCl 1806-20. Columns give the Star ID, J-band magnitude, Ks-band magnitude,J −KS value, calibrated Brγ −Kcont color (see §5.4.6), and equivalent width(including errors; see §5.4.5) of previously identified massive stars in Cl1806-20. Negative values indicate emission. All values are from this study.Values in parenthesis indicate uncertainties in the final digit of the listedmagnitudes and colors. Where no values is given, the error ≤ 0.01 mag.
Star ID J Ks J −Ks (Brγ −Kcont)cal EW(A)
Star 10 17.38(2) 11.55 5.83 0.01 2 ± 1Star B 17.84(2) 10.66 7.19 0.11 21 ± 1Star 22 17.29(2) 12.25 5.04 -0.20 -46 ± 1LBV 1806-20 13.45(2) 8.56 4.89 -0.02 -4 ± 2Star 11 16.93 11.98 4.95 0.03 5 ± 1Star 7 16.47 11.94 4.54 0.03 5 ± 1Star 4 16.20 11.21(2) 4.99 0.10(3) 21 ± 5Star C 16.48 10.89 5.60 0.05 10 ± 1Star D 15.91 11.05 4.87 0.06(2) 12 ± 3
Table 5-3. Select spectral lines in LBV 1806-20. Columns give the Line ID, restwavelength in µm and EWs from the May 2004 and July 2005 IRTF runs.
Line ID Wavelength EW (May 2004) EW (July 2005)(µm) (A) (A)
HeI 2.0581 0.15 ± 0.57 1.05 ± 0.6Fe II 2.0888 0.97 ± 0.87 1.71 ± 1.00Mg II 2.1368 2.28 ± 0.90 2.77 ± 1.13Mg II 2.1432 1.62 ± 0.80 1.72 ± 1.20Brγ 2.1655 7.95 ± 0.45 9.13 ± 0.58
165
Table 5-4. Measured magnitudes, colors, and Brγ absorption for potential OBI Stars in Cl1806-20. Columns give Star ID, J-band magnitude, Ks-band magnitude,J −KS color, corrected Brγ - Kcont (see §5.4.6) Equivalent Width, and distancefrom SGR 1806-20. Positive values for EW indicate absorption.Values inparenthesis after J −Ks and Brγ - Kcont indicate uncertainties in the last digitof the magnitude.
Star ID J Ks J −Ks (Brγ-Kcont)cal EW Distance(A) ′′
6611 18.15 13.55 4.61(3) 0.04 8 ± 2 3.26547 16.61 11.84 4.77(5) 0.14(3) 25 ± 5 6.26451 17.93 13.15 4.78(4) 0.04 7 ± 2 8.77268 18.40 13.07 5.33(5) 0.05 9 ± 1 24.86556 18.51 13.35 5.16(2) 0.02 3 ± 1 26.66091 18.40 13.65 4.75(4) 0.04 7 ± 2 40.76526 17.91 13.08 4.83(2) 0.02 5 ± 1 47.14752 17.63 12.95 4.68(3) 0.03 5 ± 1 99.73971 18.48 13.16 5.33(2) 0.02 4 ± 2 121.0
166
Figure 5-1. Histogram showing the number of stars at each calibrated broad-bandmagnitude A) Ks-band B) J-band. The bin sizes are 0.25 mag. Note theturnover at J ∼ 18 and Ks ∼ 16, indicating completeness (within the limits ofcrowding and confusion).
167
Figure 5-2. Histograms showing the number of stars at each Ks magnitude for theKs-band catalog (solid) and the 3-band catalog (dotted). The bin sizes are 0.5mag. Note the turnover at Ks ∼ 16 for the Ks-band catalog and Ks ∼ 15 forthe 3-band catalog.
168
Figure 5-3. Four histograms showing number of stars in each magnitude bin vs.magnitude. The two histograms on the right show number of stars in eachmagnitude bin verses Kcont (solid) and Brγ (dashed) instrumental magnitudes.The two histograms on the left show number of stars in each magnitude binversus Ks-band magnitude for the Ks-band catalog (dash-dot) and 3-bandcatalog (dotted). I note that the similar shapes for the 3-band, Kcont, and Brγhistograms indicates that the narrow-band data are offset from the Ks-banddata by ∼ 5 magnitudes.
169
Figure 5-4. Color-color diagram of stars in my 4-band broad- and narrow- band catalogwithin a 6′ × 6′ subfield centered on SGR 1806-20. The vertical axis is flippedsuch that negative Brγ −Kcont values, which indicate Brγ absorption, appearat the top half of the graph. Objects with previously confirmed spectral typesare noted; OBI stars are marked as triangles, WR stars as squares and theLBV as a circle. Color denotes projected distance (in ′′) from SGR 1806-20,with purple indicating objects closer to the SGR and yellow indicating objectsfurther away in 2-D space. The solid black line is the narrow-band zeropoint(see §5.4.6) with m = 0.028 and b = -0.137. Bluer stars on the left of the plotwith more scatter in their narrow-band colors are at the edge of my detectionlimits and have more Poisson noise.
170
Figure 5-5. Same plot as Fig 5-6 showing all 4-band catalog objects with Ks < 14magnitude. The candidate OBI stars are marked with inverted triangles (see§5.6.1).
171
Figure 5-6. A 1.3′ × 1.5′ region of my Ks-band image centered on Cl 1806-20. Previouslydiscovered massive stars and newly identified OBI stars detected by thissurvey are indicated – OBI stars are marked as triangles, new candidate OBIstars as inverted triangles, WR stars as squares and the LBV as a circle. Twonew OBI candidates (one ∼ 1.5′ to the northwest and another ∼ 1.5′ to thesouthwest of the cluster) are not shown.
172
Figure 5-7. Spectra of LBV 1806-20 obtained with IRTF. The top spectrum was obtainedin May 2004 and the bottom spectrum in July 2005. Each has beennormalized by the last pixel and offset from each other by a constant forclarity.
173
CHAPTER 6THE SEARCH FOR MASSIVE STARS AROUND SGRS 1900+14, 1627-41, AND
0526-66
6.1 Introduction
In Chapter 5, I detailed a method to search for massive stars around SGR 1806-20
using narrow-band imaging. In this chapter, I apply this technique to the three remaining
Soft Gamma Repeaters: SGR 1900+14, SGR 0526-66, and SGR 1627-41. I first review
previous attempts to search for clusters around each magnetar. Then, I outline my new
observations, data reduction, and photometry, specifically noting where my procedure
departed from that used for SGR 1806-20. Finally, I present my preliminary results and
discuss potential follow-up observations.
6.1.1 SGR 1900+14
While more than 50 papers written in the the last ten years describe multi-wavelength
studies of SGR 1900+14, fewer than five focus on the magnetar’s environment. This
relative paucity of observations has led to controversy over the membership of SGR
1900+14 in a nearby embedded cluster. First discovered by Vrba et al. (2000), the cluster
consists of 11 stars hidden in the glare of two previously known M5 supergiants. Using
I- and K- band photometry and assuming the same distance as that of the supergiants,
they derived absolute magnitudes for the cluster stars and constructed a color-magnitude
diagram. Their locations on the diagram suggests a supergiant or giant classification for
many of the cluster members. Citing the small probability of finding a cluster of massive
stars 12′′ from a rare compact object merely by chance, the authors concluded that the
magnetar was a member of the cluster.
Problems with this picture arose when Kaplan et al. (2002) compared the reported
distance and reddening of the cluster versus those of the magnetar. The hydrogen column
density of the magnetar as derived from X-ray spectra yields a distance of 5 kpc and a
reddening of AV ∼ 12.8 ± 0.8 mag Hurley et al. (1996); Vrba et al. (1996). On the other
hand, the distance to the cluster as determined by photometry of the supergiants is ∼ 12
174
- 15 kpc and the reddening AV ∼ 19.2 ± 1 mag (Vrba et al. 2000). This would suggest
that the SGR is not a member of the cluster and that the superposition of the embedded
cluster and magnetar is, in fact, coincidence.
This situation remained unresolved until very recently, when Wachter et al. (2008)
showed that both the cluster and the SGR are embedded in a mid-IR ring or shell, most
likely launched from the SGR and currently powered by the cluster stars. This is the
strongest evidence to date that the magnetar is a cluster member. Furthermore, the
authors calculate that reddening to the supergiants derived from 2MASS images is AV ∼10.4 - 14.4 mag; consistent with the reddening toward the SGR. One explanation for the
spread in the values of the reddening could be patchy extinction; this is particularly likely
in a field with local IR emission.
Besides lingering uncertainties in the distances and reddenings, the nature and extent
of the cluster has not been fully explored. To date, no one has performed spectroscopy
to confirm the spectral types of the candidate massive cluster stars discovered by Vrba
et al. (2000). Furthermore, the size of the MIR emission cloud enshrouding the massive
stars and magnetar points to a larger cluster than that reported by Vrba et al. (2000),
suggesting the existence of undiscovered massive stars Wachter et al. (2008). With the
opportunity to both study previously identified OBI candidates and potentially uncover
a hidden population of massive stars, 1900+14 offers an ideal opportunity to apply the
method detailed in Chapter 5.
Using a narrow-band imaging survey, I will [1] probe the field surrounding SGR
1900+14 to look for undiscovered massive cluster members [2] examine the J − Ks value
of potential cluster stars to derive the reddening to the cluster, [3] examine the Brγ
narrow-band excess of candidate massive stars identified by Vrba et al. (2000).
6.1.2 SGR 1627-41
Discovered in 1999, SGR 1627-41 is the most recent addition to the Soft Gamma
Repeater class (Woods et al. 1999). Located at a distance of 11.0 ± 0.3 kpc, it is the most
175
heavily reddened of all the SGRs with AV ∼ 43 mag (Corbel et al. 1999). SGR 1627-41
is also the least studied of the SGRs; no papers concerning the environment surrounding
SGR 1627-41 exist in the literature. Thus, my narrow-band observations of this field are
the first effort to search for a cluster associated with the SGR.
6.1.3 SGR 0526-66
Located in the LMC, SGR 0526-66 was the first magnetar to burst onto the scene in
1979, emitting a giant flare with a peak luminosity > 1044 ergs s−1 (Mazets et al. 1979).
Although centrally located in supernova remnant N49, the magnetar’s association with the
remnant is questionable; Kaplan et al. (2001) suggest an age of 1000 yr for the SGR, while
the SNR is most likely 5000 - 16000 years old (Shull 1983; Vancura et al. 1992).
While no massive cluster appears hidden by the remnant, Klose et al. (2004) report
the existence of a massive cluster, SL463, 2′ north of the SGR. Using H- and K- band
photometry, the authors construct a color-magnitude diagram and find main sequence OB
spectral types for ∼ 30 potential cluster stars. They also derive a reddening of AV ∼ 3.2
mag to SL463. Although separated by a 30 pc projected distance, the authors suggest that
SL463 is most likely associated with the magnetar, arguing that the ejection of either the
SGR or SGR progenitor from the cluster can account for the magnetar’s current location.
Currently, no follow-up observations of the findings in Klose et al. (2004) exist in
the literature. My narrow-band observations of the field surrounding SGR 0526-66 are
therefore the first attempt to locate OBI supergiants or emission-line stars enshrouded
in the cluster or located in the field between the cluster and the magnetar. Finding
these objects would make the cluster’s stellar content more analogous to Cl 1806-20 and
possibly strengthen the proposed association between the cluster and the magnetar.
6.2 Application of the Narrow-band Imaging Technique
In §5.2, I extensively detail a technique to search for massive members of Cl 1806-20
using narrow-band imaging. In this chapter, I apply this technique to SGR 1900+14,
SGR 0526-66, and SGR 1627-41. In the case of SGR 1900+14, also observed with WIRC,
176
I follow the exact same procedure that I used for SGR 1806-20. However my method is
slightly different for SGR 1627-41 and SGR 0526-66 , which I observed with the InfraRed
Side Port Imager (ISPI) on the CTIO 4-meter telescope. ISPI has dedicated continuum
filters with central wavelengths very close to the central wavelengths of the “science”
bands: the Kcont filter for HeI 2.06µm is centered at 2.03µm while Kcont for Brγ 2.16µm is
centered at 2.14µm. This small difference in the central wavelengths between the science
band and its continuum results in a negligible difference in the dust penetration between
the two narrow-band images.
Since the narrow-band color is no longer a function of reddening, it was not necessary
to obtain J or Ks broad-band images of either of the fields surrounding these clusters.
Instead, I will examine the Brγ −K2.14 color versus the K2.14 magnitude for each star in
the field. Stars with no Brγ emission or absorption should fall on a horizontal line with a
Brγ −K2.14 value equal to the differences in the transmission between the two filters. Any
stars with excess or absorption will fall above or below this “zerpoint” line. I will use the
distance from the line to calculate the EWs of any absorption or emission detected.
I will also examine the HeI − K2.03 versus K2.03 color- magnitude diagram and look
for stars with HeI excess or absorption. Since some types of massive stars display both
Brγ and HeI features in their spectra (Hanson et al. 1996), I can correlate my results to
see which stars have emission or absorption in both lines, objects in this category will be
strong candidates for follow-up spectroscopy. I discuss the advantages and disadvantages
of this technique in §6.4.2
6.3 Observations and Data Reduction
On 2005 August 26-27, I used the Wide Field Infrared Camera (Wilson et al. 2003,
WIRC) on the Palomar 200” telescope to obtain J , Ks, 2.16µm Brγ, and 2.27µm Kcont
images of an 8.7′ × 8.7′ region around SGR 1900+14. Applying standard techniques
for near-infrared (NIR) observing, I used a random 9-point dither pattern with < 30′′
separation between each image to obtain the Ks, K2.27, and Brγ data. I took three 30
177
second images at each position, resulting in 27 frames. The total exposure times were ∼8 minutes in Brγ and K2.27 and ∼ 90 seconds in Ks. Similarly, I obtained 3 exposures at
5 random dither positions with < 30′′ separation between images, resulting in a 65 second
exposure in the J-band.
On 2005 March 12-13, I used the Infrared Side Port Imager (ISPI) on the Cerro
Tololo Inter-American Observatory (CTIO) 4 meter telescope to obtain 2.16µm Brγ,
2.14µm Kcont, 2.06µm HeI, and 2.03µm Kcont images of 10.5′ × 10.5′ regions around SGR
1627-41 and SGR 0526-66.
For SGR 1627-20, I used a random 9-point dither pattern with < 30′′ separation
between each image. Again, I took three 30 second images at each pointing. I then moved
the telescope slightly (< 15′′) and repeated this procedure 4 - 5 times using a different
random dither pattern. This resulted in a total of 135 images in the HeI and K2.03 bands
and 162 images in the Brγ and K2.14. After examining the images and discarding any with
obvious defects caused by wind-shake, etc. the total exposure times were ∼ 80 minutes in
Brγ and K2.14 and ∼ 70 minutes in K2.03 and HeI.
For SGR 0526-66, I used a random 11-point dither and took one 100s exposure in
each position for Brγ and K2.14. After discarding images with defects, the total exposure
time was ∼ 25 minutes in each band. For the K2.03 and HeI observations, I returned to
the 9-point dither pattern described above. The total exposure times were ∼ 35 minutes
in K2.03 and HeI.
As described in Chapter 5, I reduced all data using FATBOY (Warner et al.in
preparation), which performed standard data reduction tasks, including dark and
sky subtracting, flat fielding, and combining of the dithered images, to produce final
reduced science frames. I performed one additional step on my SGR 1627-41 observations:
distortion correction using the IRAF package MSCTPEAK. I explore the reasons for this
in §6.4.2.1.
178
I then performed astrometry on these images using TFINDER in IRAF. For each
field, I used 500 stars to tie the four science frames to the 2MASS K-band image of the
region. This yielded excellent astrometric ties between the 2MASS images and each of the
four science frames and between the four science frames themselves. I discuss the results
of the astrometric frame-tie in §6.4.
6.4 Analysis
Here, I present a summary of the analysis of the three SGRs discussed in this chapter,
offering details only when my methods deviate from those described in Chapter 5. In
summary, I performed PSF photometry using DAOPHOT on the images in each filter
resulting in four photometric catalogs per magnetar. I matched stars across bands using
TOPCAT. Then, I identified “drop-outs”– sources that appear in one band, but are
absent from one or more of the remaining bands. I carefully checked to ensure that these
missing objects were not signs of mismatched tables or other major systematic errors that
would affect my results. Once I was satisfied that my catalogs were correct, I narrowed
my search area by applying magnitude and field size constraints. Finally, using the error
output by DAOPHOT and the values of Brγ −Kcont and HeI −K2.03 computed from the
catalogs, I calculated the equivalent widths of stars in the field surrounding the SGR.
6.4.1 Analysis of 1900+14
6.4.1.1 Photometry
I performed PSF photometry on 9654, 11396, 10861, and 9366 stars in the SGR
1900+14 Brγ, K2.27, Ks-band, and J-band images using DAOPHOT and ALLSTAR.
Using the standard error output by ALLSTAR, I calculated the median error in each
band. I found errors of 0.07, 0.04, 0.05, and 0.05 mag in the J-, Ks-, K2.27-, and Brγ-
bands respectively. I discuss the errors associated with this technique in greater detail
in §6.4.1.5. Next, I calibrated the J and Ks magnitudes for my sources using 2-MASS
photometry. I chose 20 isolated stars with a range of magnitudes from the 2MASS image
of my field and matched them with their counterparts in my science frames. Starting
179
with the Ks-band, I calculated the difference between my ALLSTAR magnitude and the
2MASS magnitude for each star, found the median value of the offset, and applied it to
the remainder of my Ks-band data. I repeated this process for the J-band photometry.
6.4.1.2 Star matching
Upon completing the astrometry for SGR 1900+14 using TFINDER in IRAF, I chose
50 isotropically distributed stars in one frame, found their counterparts by eye in the other
3 frames, and recorded their coordinates on all four images. I then calculated the average
offset between these positions, which is a measure of the astrometric accuracy between
the frames. I found an offset 0.09 ± 0.08′′ between the Brγ and K2.27 frames, 0.13 ± 0.9′′
between the J and Ks images, and 0.09 ± 0.08′′ between Brγ and Ks. This was in good
agreement with the errors reported by IRAF between each frame and the 2MASS image
used to complete the astrometry.
Using TOPCAT, I matched the four photometric catalogs for SGR 1900+14. I used a
0.4′′ radius to match K2.27 (RA, Dec) to Brγ (RA, Dec) and 0.4 ′′ radius to match J-band
(RA, Dec) to Ks-band (RA, Dec) for all detected stars. I then matched the two tables
to each other, using Brγ and Ks-band coordinates and a 0.4′′ radius. I calculated these
radii from the astrometric errors using the error radius plus 3σ. Optimally, I would have
matched all four bands simultaneously, however TOPCAT was unable to perform this
match directly given the size of the data sets.
I found 5274 J-band counterparts to Ks-band sources, 7866 K2.27 counterparts to
Brγ sources and 4574 sources with matches across all four bands. I discuss these statistics
further in §6.4.1.3.
6.4.1.3 Limiting magnitudes and match drop-outs
Using the 2MASS calibrated photometric catalogs, I determined the limiting
magnitudes for the J- and Ks- band images of the field surrounding SGR 1900+14.
In Figure 6-1 I present histograms of both Ks-band magnitude and J-band magnitude
versus number of stars at each magnitude. These distributions peak at Ks ∼ 15.5 mag
180
and J ∼ 17.5 mag. As potential OBI supergiants in the cluster will fall between ∼ 13 - 16
mag in the J-band (Vrba et al. 2000), I confirmed that my J-band photometric limits are
deep enough to observe known cluster members in the J-band.
To confirm that my Ks-band photometric limit is adequate to detect potential
massive cluster stars, I estimated Ks-band magnitudes using known J-band magnitudes
of candidate OBI stars in the embedded cluster. Given that the values for the reddening
toward the cluster range from AV ∼ 10.4 - 19.2 (Vrba et al. 2000; Wachter et al. 2008), I
expect a minimum J − Ks value of approximately 2 magnitudes for stars in the cluster.
Assuming an intrinsic J − Ks ∼ 0 for OBI stars (Cox 2000), I anticipate massive cluster
members to be at least 2 magnitudes brighter in the Ks-band than the J-band. Therefore,
the most conservative values for the reddening predict that the massive stars in the cluster
will fall between ∼ 11 - 14 mag in the Ks-band, well within my photometric limits.
As with Cl 1806-20, I found a considerable number of drop-outs each time I matched
catalogs. Drop-outs are defined as sources that appear in one band, but are absent from
one or more of the remaining bands. In §5.4.3, I discuss a variety of reasons for drop-outs
in the Cl 1806-20 data. I found that drop-outs are most likely [1] stars at the field edges
missing from one or more images due to small variations in the central positions of the
field-of-view between frames or [2] very faint stars at the detection limits of one or both
bands. In both cases these are unlikely to be the bright supergiants or emission-line stars
that I am searching for here. While in Cl 1806-20, a small percentage of drop-outs were
potentially interesting objects in the most crowded regions of the field, it was obvious that
adjusting detection thresholds in DAOPHOT to find every star in the frame would simply
result in a large number of incorrect matches and spurious detections of background noise.
After completing a similarly extensive review of the drop-outs between each band in the
SGR 1900+14 catalogs, I come to similar conclusions, which I detail below.
The largest number of drop-outs occurred between the J- and KS- band match.
Finding Ks-band sources with no J-band counterpart was unsurprising; since longer
181
wavelengths are more efficient at penetrating dust I expected many of the Ks stars to be
highly obscured and thus undetected in the J-band. However, I noted that a number of
(∼ 4000) J-band sources were missing from the J- and KS- band match. After examining
both the J- and Ks- band limiting magnitudes in each individual catalog compared to the
2-band match, I found that the majority of these drop-outs were dim J-band stars with
no K-band counterparts. These stars, with a median J-band magnitude = 18.5 are > 2
magnitudes dimmer than any massive star currently known in the cluster. Thus, these
drop-outs should not affect my results.
I then inspected the drop-outs in my Brγ and K2.27 band match. Given the slightly
longer central wavelength of the K2.27 filter, I expected to find some unmatched faint
K2.27 sources. However, I also found ∼ 2000 Brγ dropouts. After analyzing histograms
of narrow-band magnitudes versus star counts and images of the stars on each frame, I
found that the majority of the unmatched sources are faint objects at the detection limits
of both filters. Thus, slight difference in limiting magnitudes accounted for the majority of
missing counterparts between the two narrow-bands.
Finally, I looked for drop-outs in my 4-band match. I found ∼ 700 drop-outs between
the J- to Ks- band catalog to the final 4-band match. As with the Cl 1806-20 data,
I concluded that the majority of these are caused by a drop in the sensitivity of my
narrow-band data due to differing integration time; my narrow-band observations were too
short to reach the same limiting magnitude as my broad-band imaging. However, given
the expected brightness of potential massive cluster stars (Ks < 14) as discussed in the
beginning of this section, this magnitude limit for my narrow-band imaging should not
adversely affect my results.
6.4.1.4 Field selection
My observations of the three SGRs discussed in this chapter encompass 8 - 10′ fields
centered on the magnetars. However, in the case of SGR 1900+14, where a large massive
182
cluster has already been detected near the magnetar, I can narrow my search for massive
stars by considering a subregion of my data.
I chose the subregion around 1900+14 using the estimates of the Galactic massive
star clusters whose stellar contents, ages, and (potentially) masses are similar to Cl
1806-20. These were discussed in detail in §5.4.4. At 12 kpc, the “middle” distance
estimate to the cluster associated with SGR 1900+14, a search area with a 16pc diameter
corresponds to a ∼ 2.5′ radius centered on the magnetar. If the SGR is at the ∼ 5 kpc
distance suggested by Hurley et al. (1996), my search area would correspond to a 7 pc
diameter search area; still large enough to contain all but the most extreme outliers.
I present the color-color diagram for stars within a 5′ diameter of SGR 1900+14 in
Figure 6-2. I note that the J −Ks color is a distance and reddening indicator; objects on
the left of the graph are bluer and therefore most likely located in the foreground, while
objects to the far right are heavily reddened. Since AK = 0.112AV and AJ = 0.282AV , I
can use the conflicting reddening values to help constrain potential search areas for the
cluster. Using an AV = 10.4 - 14.4 for SGR 1900+14, (Wachter et al. 2008) I find J −Ks
= 1.8 - 2.5 mag. For an AV = 19.2 ± 1 for SGR 1900+14 (Vrba et al. 2000) I find a
J −Ks = 3.3 ± 0.2 mag. I discuss my results in §6.5.1.
6.4.1.5 Calculating errors and equivalent widths
Using my newly created four-band catalog and the errors output by ALLSTAR, I
calculated the error in the Brγ −Kcont narrow-band color for every star in the 5′ diameter
subregion. I found an average error of 0.04 mag corresponding to ∼ 4% photometry; the
limit necessary to allow for 3σ detections of modest Brγ emission and absorption lines
(§5.2). This error is lower than the median errors calculated using the entire field. As
with Cl 1806-20, the elimination of dim stars (a by-product of the catalog matching) and
slightly distorted stars (by the application of a subfield) accounted for this decrease.
Once I determined that my global photometry satisfied the 4% criterion, I assessed
my error on a more source-specific scale. I computed the errors in the J − Ks and
183
Brγ − Kcont colors for every star in my 4-band catalog using the errors output by
ALLSTAR. If a particular source displays emission or absorption but a large error in
either color, I can use these errors to evaluate whether it is still a good candidate for
follow-up spectroscopy. I discuss this in more detail in the sections below.
In Figure 6-2, I note a linear correlation between the J − Ks color and Brγ − K2.27
color of sources in the field. Since longer wavelengths penetrate dust more efficiently, the
K2.27 filter, centered at a longer wavelength than the Brγ filter, systematically detected
more flux, even if no intrinsic excess or absorption existed. The size of this effect depends
on the extinction along the line-of-sight; the slope of the line is therefore an indicator of
the varying reddenings and distances of objects in the field-of-view.
Using TOPCAT, I fit a line with slope (m) = 0.021 and y intercept (b) = -0.143 to
the data. This line is a “zeropoint”; it defines the expected value of Brγ −Kcont for a star
with neither Brγ absorption nor emission. Using this linear equation, the Brγ − Kcont
values of stars in my catalog, errors in the narrow-band and broad-band color, and the
technique described in §5.4.6, I calculated the equivalent widths and associated error for
potential massive star candidates. I present these results in §6.5.1
6.4.2 Analysis of 1627-41
6.4.2.1 Departures from the standard technique: distortion correction andfield selection
In §6.2, I outlined a departure from the original method used to search for massive
stars around SGR 1900+14 and SGR 1806-20. While this technique continues to use
narrow-band colors to find emission or absorption indicative of massive stars, I use two
narrow-band “color-magnitude” diagrams: HeI−K2.03 versus K2.03 and Brγ−K2.14 versus
K2.14 instead of a 4-band “color-color” diagram, to search for massive stars.
One advantage of this method is the ability to correlate the results; since some
massive stars exhibit both HeI and Brγ spectral features (Hanson et al. 1996), evidence
of emission or absorption in both the HeI − K2.03 and Brγ − K2.14 color-magnitude
184
diagrams may indicate a strong candidate for follow-up spectroscopy. This is, however, not
a requirement for establishing massive star candidacy; many massive stars do not exhibit
features in both lines.
Unlike the other magnetars in my study, I search for massive stars across the entire
10.4′ × 10.4′ field surrounding SGR 1627-41. Since this is the first attempt to find massive
stars associated with the magnetar, I do not want to exclude any region from my analysis.
This is particularly important given possible formation scenarios of magnetars; it is not
unreasonable for a supernova associated with the death of a massive star to eject the
resulting compact object from its natal cluster.
The decision to survey the entire ISPI frame resulted in an extra step in the reduction
of the SGR 1627-41: the application of a distortion correction. Upon examining my
unreduced ISPI data, I noted a significant distortion of stellar PSFs in the outer regions
of my images. This distortion not only affects the location of the central pixel in each
star, limiting my ability to perform accurate frame-ties, it increases errors in the PSF
photometry.
I completed a distortion correction for each of my narrow-band filters using the
TFINDER package in IRAF. I did not perform this correction on my final image; instead
I used one of the flat-fielded, dark- and sky- subtracted frames produced by FATBOY. I
then applied this correction to approximately 150 reduced frames per filter before drizzling
them together to create the final science frame. This resulted in tighter stellar profiles in
every dithered image, ensuring that the PSF in the final frame was not artificially inflated
by the superposition of many irregularly shaped distorted stars.
6.4.2.2 Photometry and band matching
I performed PSF photometry on 23284, 25487, 17371, and 19268 stars in the SGR
1627-41 Brγ, K2.14, HeI, and K2.03 band images using DAOPHOT and ALLSTAR. I
calculated median errors of 0.05, 0.05, 0.04, and 0.04 mag across the Brγ, K2.14, HeI, and
K2.03 bands respectively.
185
Upon completing astrometry for SGR 1627-41 using TFINDER in IRAF, I calculated
the astrometric error as described in §6.4.1.2. I found an average offset of 0.18 ± 0.09
between the Brγ and K2.14 frames, 0.12 ± 0.07 between the HeI and K2.03 images, again
in good agreement with the errors reported by IRAF between each frame and the 2MASS
image used to complete the astrometry.
Using TOPCAT, I matched the photometric catalogs for SGR 1627-41. I used a 0.5′′
radius to match K2.14 (RA, Dec) to Brγ (RA, Dec) and 0.4 ′′ radius HeI (RA, Dec) to
K2.03-band (RA, Dec) for all detected stars. I calculated these radii from my astrometric
errors using the error radius plus 3σ. Since I was not creating a color-color plot, I did
not need a four-band match for my primary analysis, however, I did match the two
narrow-band tables to simplify my search for stars with both HeI and Brγ features (§6.2.
I used a radius of 0.4′′ for this match. I found 14620 K2.03 counterparts to HeI sources,
19043 K2.14 counterparts to Brγ sources, and 13507 sources with matches across all four
bands. I discuss these statistics further in §6.4.2.3.
I present my HeI − K2.03 versus K2.03 and Brγ − K2.14 versus K2.14 diagrams in
Figures 6-3 and 6-4.
6.4.2.3 Limiting magnitudes and match drop-outs
One disadvantage of eschewing J- and Ks- band photometry in favor of a second
set of narrow-band observations is the lack of a universally available photometric catalog
to calibrate narrow-band data. For my analyses of SGR 1806-20 and SGR 1900+14, I
calibrated my instrumental broad-band magnitudes using 2MASS. I then used these
results to estimate the limiting magnitudes of my images and compared them to the
magnitudes of potential cluster stars. Thus, I could assure that my photometry was deep
enough to observe any potential massive stars. The uncalibrated instrumental magnitudes
output by DAOPHOT for the stars surrounding SGR 1627-41 do not allow for such an
assessment.
186
To address this issue, I compared my narrow-band data with 2MASS broad-band
images. Using the distance to and reddening of the cluster, I calculate that a typical OBI
star with an absolute magnitude MK ∼ 6 mag will have an apparent magnitude mk ∼14 mag. I found 25 stars in my field with a Ks-band magnitude ∼ 14.5 in the 2MASS
catalog. I then located these stars in my 4-band match and recorded the magnitude of
each object in all four bands. I found that the mean instrumental magnitudes of ∼ 14.5
magnitude 2MASS stars in my Brγ, K2.14, HeI, and K2.03-images are 21.0, 20.8, 21.2, and
21.0 respectively. This is brighter than the 22.0 mag limiting instrumental magnitude
of all four bands. While this calibration is by no means extensive, it suggests that my
observations are deep enough to observe most massive stars around SGR 1627-41, should
they exist.
I then examined drop-outs within my two narrow-band catalogs. I determined that
in most cases, the drop-outs resulted from a slight difference in the limiting magnitudes of
the narrow-band filters. The vast majority of drop-outs are very faint, beyond the limiting
instrumental magnitude in each frame. Considering the magnitude arguments established
above, it is unlikely that these drop-outs will affect my results.
6.4.2.4 Photometric errors and equivalent width calculations
The median error in both the Brγ − K2.14 and HeI − K2.03 colors for SGR 1627-41
is 0.06 mag, slightly larger than the errors reported for other fields in this study. I
hypothesize that my initial decision to survey the entire field surrounding SGR 1627-41
caused the elevation in my photometric errors. I noted that distortion effects, which could
result in poor PSF fits and therefore larger errors, persisted at the edges of the final frame
despite my distortion correction. I discuss this issue in greater detail in §6.5.2.
These errors may also reflect the different technique applied to the field surrounding
SGR 1627-41. Since this method uses Brγ − K2.14 versus K2.14 and HeI − K2.03 versus
K2.03 diagrams, I produced two 2-band catalogs as opposed to one 4-band match. The
4-band catalogs, used to create “color-color” diagrams for SGR 1900+14 and SGR
187
1806-20, excluded thousands of the dimmest sources in every band. The inclusion of these
faint objects in the 2-band catalogs used to calculate the color error for the SGR 1627-41
data increased the total Poisson noise, potentially resulting in the 0.06 mag median errors.
Given that my global errors are larger than the 4% criterion established to yield 3 - 4
σ detections of moderate narrow-band emission and absorption lines, I placed more weight
on the Brγ −K2.14 and HeI −K2.03 errors that I calculated for each source. As with the
other clusters, I ultimately used these values to determine whether a star was a viable
candidate for follow-up spectroscopy.
Finally, I calculated the EWs of stars in my color-magnitude diagrams. Unlike the
color-color diagrams used for SGR 1806-20 and SGR 1900+14, the “zeropoint” line in
Figures 6-3 and 6-4 are independent of reddening and therefore have a slope (m) ∼ 0.
Since the majority of sources do not have Brγ or HeI spectral features, their narrow-band
colors should be zero. However, an offset, (Y) indicative of the transmission differences
between the filters, is evident upon inspection of both color-magnitude diagrams. To find
this offset, I calculated the median value of each of the narrow-band colors for all objects
in the field. I found Y = 0.17 for Brγ − K2.14 and Y = 0.24 for HeI − K2.03. To find
the calibrated color of my sources, I subtracted the corresponding value of Y from the
HeI −K2.03 or Brγ −K2.14 value of each star. I then calculated the EWs and errors using
the techniques described in §5.4.6. I present these results in §6.5.1
6.4.3 Analysis of 0526-66
6.4.3.1 Photometry and band matching
I performed PSF photometry on 1473, 1604, 1097, and 984 stars in the SGR 0526-66
Brγ, K2.14, HeI, and K2.03 band images using DAOPHOT and ALLSTAR. I calculated
median errors of 0.1, 0.1, 0.15, and 0.15 mag across the Brγ, K2.14, HeI, and K2.03 bands
respectively. I attribute these large errors to the overabundance of very faint stars in the
field, which I discuss further in §6.4.3.3.
188
Upon completing astrometry for SGR 0526-66 using TFINDER in IRAF, I calculated
the astrometric error as described in §6.4.1.2. I found an average offset of 0.20 ± 0.07
between the Brγ and K2.14 frames, 0.07 ± 0.07 between the HeI and K2.03 images, again
in good agreement with the errors reported by IRAF between each frame and the 2MASS
image used to complete the astrometry.
Using TOPCAT, I matched the photometric catalogs for SGR 1627-41. I used a 0.5′′
radius to match K2.14 (RA, Dec) to Brγ (RA, Dec) and 0.3 ′′ radius HeI (RA, Dec) to
K2.03-band (RA, Dec) for all detected stars. I calculated these radii from my astrometric
errors using the error radius plus 3σ. Like SGR 1627-41, I used two color-magnitude
diagrams to analyze the stars surrounding SGR 0526-66. Since I was not creating a
color-color plot, I did not need a four-band match for my primary analysis, however, I did
match the two narrow-band tables to simplify my search for stars with both HeI and Brγ
features (§6.2. I used a radius of 0.4′′ for this match. I found 840 K2.03 counterparts to
HeI sources, 942 K2.14 counterparts to Brγ sources, and 588 sources with matches across
all four bands. I discuss these statistics further in §6.4.2.3.
I present my HeI − K2.03 versus K2.03 and Brγ − K2.14 versus K2.14 diagrams in
Figures 6-5 and 6-6
6.4.3.2 Limiting magnitudes and match drop-outs
Without J- and Ks- band photometry I could not easily determine whether the
narrow-band observations of SGR 0526-66 were deep enough to detect massive stars
around the magnetar. However, unlike SGR 1627-41, SGR 0526-66 has a potential
associated massive cluster, SL 463, with one OBI candidate. This star has a reported
K-band magnitude ∼ 12 mag (Klose et al. 2004). Thus, if my observations reach this
depth, I can be assured that I will detect other potential massive cluster stars, should they
exist.
I compared my narrow-band frames to 2MASS broad-band images to find a
rough offset between standard 2MASS Ks-band magnitudes and my instrumental
189
magnitudes. I chose 25 isolated stars in my field with a broad distribution of 2MASS
Ks- band magnitudes. I then located these stars in my 2-band catalogs and recorded
the instrumental magnitude of each object in all four bands. I found a ∼ 6.5 magnitude
offset between the 2MASS Ks- band magnitudes and my instrumental magnitudes,
approximately the same offset for the SGR 1627-41 data. This yielded an expected
narrow-band instrumental magnitude of 18.5 mag for a 12 mag star, well within the 22.5
mag limiting instrumental magnitude for the K2.03 and K2.14 bands and 23 mag limiting
instrumental magnitude for the HeI and Brγ bands.
I then examined drop-outs within my two narrow-band catalogs. Again, I determined
that in most cases, the drop-outs resulted from a slight difference in the limiting
magnitudes of the narrow-band filters. The vast majority of drop-outs are very faint,
beyond the limiting instrumental magnitude in each frame. Considering the magnitude
arguments established above, it is unlikely that these drop-outs will affect my results.
6.4.3.3 Field selection, errors, and equivalent widths
As mentioned in §6.1.3, SL463, the potential natal cluster of the SGR, is located 2′
away from SGR 0526-66. Thus, I chose a 4′ search radius centered around the SGR that
encompassed both the cluster and the magnetar. Using the errors output by ALLSTAR
for the stars in this subfield, I computed a 0.04 mag median error for both the Brγ −K2.14
and HeI − K2.03 colors. This met the 4% criterion required to detect moderate spectral
features in massive stars to a 3σ level. I then computed the errors in the J − Ks and
Brγ − Kcont colors for every star in my 4-band catalog using the errors output by
ALLSTAR.
Finally, I calculated the EWs of stars in my color-magnitude diagrams, which I
present in Figures 6-5 and 6-6. Like SGR 1627-41, both narrow-band colors are offset
from zero, representing transmission differences between the bands. To find this offset, I
calculated the median value of each of the narrow-band colors for all objects in the field.
I found Y = 0.27 for Brγ − K2.14 and Y = 0.22 for HeI − K2.03. To find the calibrated
190
color of my sources, I subtracted the corresponding value of Y from the HeI − K2.03 or
Brγ −K2.14 value of each star. I then calculated the EWs and errors using the techniques
described in §5.4.6. I present these results in §6.5.3
6.5 Preliminary Results
6.5.1 SGR 1900+14
To better identify previous cluster candidates and find new potential cluster members,
I further constrain my search to stars with Ks ≤ 14.2. I show the resulting plot in Figure
6-7. Using the most conservative estimate for the reddening, I anticipate that stars in
the cluster should have Ks-band magnitudes between ∼ 11 - 14 mag; by removing the
dimmer foreground stars, I limit contamination by foreground stars with large broad-band
photometric errors. I also exclude fainter narrow-band sources, limiting the error and
reducing the scatter around the line. Stars with Brγ excess or absorption, discussed in
detail below, are clearly labeled in Figure 6-8, a 1.5′ × 1.5′ region of my Ks-band image
centered on SGR 1900+14 .
By comparing my observations to previous results in the literature (Table 6-1), I am
better able to evaluate how well my application of this narrow-band technique worked to
identify potential massive stars around SGR 1900+14. Vrba et al. (2000) found 11 OBI
candidates in the embedded cluster associated with the magnetar. Unfortunately, the
majority of these stars lie directly next or behind one or both of the bright M5 supergiant
cluster members. While Vrba et al. (2000) is able to remove these bright objects from
their I and J -band image using PSF subtraction, this was impossible in our Ks - band
image, where the stars are completely saturated, even using the shortest exposure time.
Given these limitations, only four of the stars are not affected by saturated pixels, namely
Stars 1, 2, 3, and 4. These stars are shown by triangles in Figure 6-7.
While Star 1 shows moderate Brγ absorption (1 -4A) suggesting that it may be an
OBI star, Stars 4 and 2 are both consistent, within the error bars, with no absorption or
emission. Star 3, meanwhile, exhibits a Brγ emission with EW = -(5 - 13)A. Also, while
191
the other three stars are located at a J−Ks ∼ 2, Star 3 is at a J−Ks ∼ 2.6. One possible
explanation for this discrepancy may be that Star 3 is a dust emission WR, with warm
circumstellar dust intrinsically reddening the star and producing a slightly redder color
than other cluster members. In this case, the strength of the emission may be masked; the
line could be even stronger than predicted by the narrow-band color.
Another option is the presence of patchy extinction along the line-of-sight to the
cluster, resulting in varying values of J − Ks for different cluster members. This
explanation might also explain the broad distribution of reddenings (Av = 10.4 - 19.2
mag) derived for the cluster. As discussed in §6.1.1, Vrba et al. (2000) suggest a reddening
Av = 19.2 ± 1 mag yielding a J −Ks = 3.2 ± 0.18 color for the cluster (assuming massive
stars with no intrinsic color term) while Wachter et al. (2008) find Av = 10.4 - 14.4 mag
or a J − Ks = 1.8 - 2.6 mag. While I observe a few stars in the J − Ks = 3.2 ± 0.18
mag regime of both Figure 6-7, the J − KS color of four of the stars identified by Vrba
et al. (2000) (shown as triangles) are more consistent with Av = 10.4 - 14.4 mag. Thus my
results support those of Wachter et al. (2008) and are also consistent with the reported
reddening of the SGR, AV ∼ 12.8 ± 0.8 mag (Vrba et al. 1996). However, it is clear that a
more thorough study of the extinction law in the vicinity of the cluster is necessary.
I then compared my J-band magnitudes for Stars 1 - 4 to other values in the
literature. While similar to the J- band magnitudes reported by Vrba et al. (2000),
they do not agree within the error bars; my observations are systematically dimmer by
0.06 - 0.14 mag. This is notable given the high precision of both sets of photometry. Slight
differences in the zeropoint used to calibrate the instrumental magnitudes or errors in the
PSF subtraction of the nearby M5 supergiants may account for these discrepancies.
Before exploring potential massive star candidates in the field surrounding SGR
1900+14, I compare my results with those from SGR 1806-20. I immediately note
significant differences between the color-color diagram of SGR 1900+14 (Figure 6-7) and
SGR 1806-20 (Figure 5-4). In Figure 5-4, the region surrounding J − Ks = 5, the color
192
of the cluster, is sparse, making it relatively easy to pick out potential cluster candidate
using reddening constraints alone. In6-7, the scatter in the Brγ − K2.27 color at the
potential J − Ks colors of SGR 1900+14 hinder my ability to easily identify potential
massive stars.
To solve this problem, I applied an additional constraint to my search; I identified
all stars with Ks-band magnitudes > 14.2 within a 45′′ radius of the SGR and examined
their narrow-band colors. I chose this radius based on the size of the mid-infrared ring
discovered around SGR 1900+14 by Wachter et al. (2008). While this method may
exclude any massive cluster members ejected from the cluster, it offers a better chance to
locate massive stars in the cluster’s 1 - 2 pc core should they exist.
I show the final color-color diagram in Figure 6-9. Objects within a 45′′ radius of the
SGR with Ks-band magnitudes > 14.2 that have EW consistent with either Brγ emission
or absorption after taking into account the errors, are marked with inverted triangles.
These results are presented in Table 6-2. These objects, potential OBI and emission line
stars, are the most interesting candidates for follow-up spectroscopy and may represent
the bulk of previously undiscovered massive star population in this cluster.
6.5.2 SGR 1627-41
Figures 6-3 and 6-4 show the HeI −K2.03 versus K2.03 and Brγ −K2.14 versus K2.14
diagrams for the stars surrounding SGR 1627-41. Since there have been no prior attempts
to locate a cluster associated with this magnetar, I do not have previous data with which
to compare my observations.
In both color-magnitudes diagrams, there is an obvious correlation between
“magnitude” and error in the narrow-band color; the error increases with magnitude,
resulting in a larger spread of the data away from the “zeropoint” line. This spread,
combined with the sheer number of stars on the plot, made it impossible, to identify an
obvious cluster of stars with either HeI − K2.03 or Brγ − K2.14 emission or absorption.
While my initial intention was to avoid focusing on any one region of the field, I
193
hypothesized that, like SGR 1900+14, limiting my search to a small area surrounding
the SGR might offer the best opportunity to locate massive stars associated with the
SGR.
I located all stars within a 45′′ radius from the SGR corresponding to a ∼ 2 pc
search radius at the distance of the magnetar. First, I calculated the Brγ − K2.14 and
HeI −K2.03 colors of all the stars in this subregion, taking into account the error bars. I
found a number of objects with only Brγ emission or absorption; these stars are marked
with diamonds on Figure 6-3. I also found a few candidates with only HeI emission ; these
objects are denoted as squares. Objects with emission or absorption in both HeI and Brγ
are denoted as triangles on both plots. I list all potential massive star candidates and
their EWs in Table 3. Stars with Brγ and/or HeI excess or absorption, are clearly labeled
in Figure 6-10, a 1′ × 1′ region of my Brγ image centered on SGR 1627-41.
While my analysis of the field surrounding SGR 1627-41 succeeded in identifying
potential massive stars around the magnetar, the amount of error in the narrow-band
color raises concerns. Evidence of this problem surfaced when I was evaluating the PSF
photometry; the error in the narrow-band colors for the entire data set did not meet the
original 5% criterion discussed in the methodology section of Chapter 5.
One potential source of photometric error is the distortion correction performed
during the reduction of my observations. I noted that even after considerable correction,
distorted PSFs were still evident in the final science images, especially toward the outer
edges. While the distortion correction completed in IRAF on each sky-subtracted image
did eliminate the errors in that frame, applying this calibration to all ∼ 150 images
observed with the same filter may have not solved the distortion problem. Flexure in the
telescope and instrument can create a different distortion pattern at each hour angle; since
I took my observations over a number of hours, the distortion correction calculated from
data at the beginning of the night may not fit data observed later.
194
One solution to this problem would be to find a distortion correction for each of the
six dither runs completed per filter. I could then apply the correct distortion pattern
to each set of dithered frames and combine the results into a final science image. In the
future, I will attempt this correction in the hope of reducing the errors in my narrow-band
color. Once I reduce the scatter on my color-magnitude diagrams, I hope to further
explore the rest of the field to search for more massive star candidates.
6.5.3 SGR 0526-66
Since Klose et al. (2004) do not provide photometry or proposed spectral types for
individual cluster stars in SL463, I cannot compare my results for SGR 0526-66 on a
star-by-star basis with previous data in the literature. However, I do identify the single
OBI star reported by Klose et al. (2004) and find Brγ absorption; I estimate a Bryγ EW
= 9 - 13 A.
Upon visual inspection of Figures 6-3 and 6-4, I note the absence of an obvious
population of stars with HeI and Brγ emission or absorption. Given the ∼ 6.5 magnitude
offset between the Brγ instrumental magnitudes and broad-band Ks magnitudes
(§6.4.3.2), I expect to find potential massive cluster members with K-band magnitudes
∼ 12 mag (Klose et al. 2004) at a K2.14 ∼ 18.5 mag and K2.03 ∼ 18.5 mag. Besides
an obvious paucity of stars at this magnitude on the color-magnitude diagrams, my
calculations show that the few bright stars do not exhibit narrow-band spectral features.
To more throughly search for massive stars in this field and reduce the scatter in
the narrow-band colors, I further limited my search area to a 45′′ radius centered on the
cluster, corresponding to the cluster size estimated by (Klose et al. 2004) (see Figures 6-11
and 6-12). Additionally, I searched another 45′′ radius surrounding the SGR itself (see
Figures 6-13 and 6-14). At 50 kpc, my search areas correspond to a 1 pc cluster radius;
the typical size of a cluster core (§5.4.4).
I detected no signatures of massive stars in the search area centered on the SGR, a
result in agreement with the findings of Klose et al. (2004). However, given the presence
195
of bright NIR nebulous emission from nearby SNR N49, I can not completely rule out the
presence of massive stars. I also failed to find additionally OBI candidates in SL463. I do
detect many of the candidate main sequence OB cluster members. As expected of stars
of these spectral types, these objects show no signs of HeI or Brγ absorption or emission.
I present an image of the two search areas in Figure 6-15. A box indicates the one OBI
candidate in SL463.
Clearly, more work to establish whether SL463 is indeed the natal cluster of SGR
0526-66 is necessary. Broad-band photometry and a color-color diagram, like those created
for SGR 1900+14 and SGR 1806-20, may help illuminate the stellar population in the
field. Reducing the photometric error and culling foreground objects by identifying stars
at the reddening of the cluster should allow for a more conclusive result regarding SGR
0526-66 and its environment.
6.6 Conclusions
In this chapter, I describe my search for massive stars associated with SGR 1900+14,
SGR 0526-66, and SGR 1627-41 using a narrow-band technique successfully applied to Cl
1806-20.
1) I observed the embedded cluster associated with SGR 1900+14. While I could
not analyze many of the OBI candidates discovered by Vrba et al. (2000) due to their
proximity to two bright M5 supergiants, I successfully detected four previously identified
OBI candidates. I confirm Star 1 as an OBI candidates and suggest that Star 3 may
be an massive star with Brγ emission. My data support the reddening values suggested
by Wachter et al. (2008), Av = 10 - 14.4 mag. Furthermore, I identified a population of
candidate massive stars that may represent the bulk of the missing cluster population. I
suggest that these stars should be targeted for future spectroscopic observations.
2) Using color-magnitude diagrams, I probed the environment of SGR 1627-41 for
massive stars. While I found a number of potential massive stars with Brγ and HeI
196
emission and absorption, I noted the large photometric error associated with the data. I
discuss corrections to the data that may yield better results.
3) Using the same technique applied to SGR 1627-41, I probed two 45′′ regions:
one around cluster SL463 and another around SGR 0526-66. While I did identify the
previously discovered lone OBI candidate in SL463, I found no evidence for a large OBI or
emission-line star population. Further work to evaluate whether SL463 is indeed the birth
cluster of the magnetar and to determine the stellar content of is clearly necessary.
4) Upon comparing the technique used for SGR 1806-20 and SGR 1900+14 to that
used for SGR 1627-41 and SGR 0526-66, I find the former to be more effective. Although
this method was originally contrived to sidestep an observational difficulty – the lack of
a continuum filter at a wavelength close to the central wavelength of the science filter,
the color-color diagram proved to be more useful. In situations where the J − Ks color
of potential massive stars is known, I could focus on a specific region of the diagram to
search for potential cluster members.
5) The color-color method turns out the best results when the cluster is at a high
reddening and sharply separated from the remaining data, as was the case with Cl
1806-20. Thus, a similar analysis for Cl 1627-41, at a very large reddening might yield
better results.
197
Table 6-1. Measured magnitudes, colors, and Brγ excess or absorption for previouslyknown massive stars in Cl 1900+14. Columns give the Star ID, I-bandmagnitude, J-band magnitude from (Vrba et al. 2000) then continue withresults from this study including J-band and Ks-band magnitudes, calibratedBrγ −K2.27 color (see §5.4.6), and equivalent width (including errors; see§5.4.5). Negative values indicate emission. Stars with no photometry from thisstudy lie directly behind or next to one of two M5 supergiants in the cluster.
Star ID Ia Ja J Ks (Brγ −K2.27)cal EW(A)
Star 1 20.00 ± 0.02 13.17 ± 0.02 13.26 ± 0.01 11.22± 0.01 0.01 1 - 4Star 2 21.10 ± 0.02 14.26 ± 0.02 14.37 ± 0.02 12.37± 0.03 0.00 0Star 3 21.85 ± 0.02 14.51 ± 0.02 14.68 ± 0.02 12.12± 0.01 -0.04 -(5 -13)Star 4 22.01 ± 0.02 15.06 ± 0.04 15.20 ± 0.04 12.98± 0.04 0.03 0 a
Star 5 22.77 ± 0.03 .... ... ... ... ...Star 6 23.18 ± 0.04 .... ... ... ... ...Star 7 23.16 ± 0.05 15.70 ± 0.09 ... ... ... ...Star 8 22.85 ± 0.04 15.62 ± 0.06 ... ... ... ...Star 9 23.01 ± 0.05 15.53 ± 0.05 ... ... ... ...Star 10 23.60 ± 0.06 16.41 ± 0.07 ... ... ... ...Star 11 23.78 ± 0.06 16.12 ± 0.04 ... ... ... ...a Vrba et al. 2000b consistent with 0 within the error bars
198
Table 6-2. Measured magnitudes, colors, and Brγ emission or absorption for potentialmassive stars in the embedded cluster around SGR 1900+14. Columns giveStar ID, J-band magnitude, Ks-band magnitude, J −KS color, corrected Brγ -K2.27 (see §5.4.6) Equivalent Width, and distance from SGR 1900+14. Positivevalues for EW indicate absorption. Unless otherwise stated, errors in the J-and Ks-band magnitude are ± 0.01
Star ID J Ks J −Ks (Brγ-K2.27)cal EW Distance(A) ′′
3588 16.04 13.66 2.37 0.023 2 - 7 43.13674 15.10 13.19 1.91 0.060 10 - 13 39.93883 14.14 12.96 1.19 0.023 3 - 7 34.93970 13.75 10.88 2.87 -0.011 (-)1 - 3 23.54137 16.27 14.11 2.15 -0.029 (-)3 - 9 17.94498 14.40 11.94 2.46 -0.011 (-)1 - 4 3.44554 14.86 13.68 1.18 -0.018 (-)1 - 7 8.04586 13.95 11.59 2.36 -0.012 (-)1 - 4 22.34715 16.91± 0.03 14.12 2.79 0.026 1 - 9 18.64761 16.49± 0.03 14.04 2.46 -0.048 (-)6 - 14 45.04937 15.26 14.13 1.13 0.03 2 - 9 36.05031 14.89 12.67 2.22 0.025 4 - 6 31.45095 16.18± 0.02 14.04 2.14 0.022 1 - 8 38.85180 15.05 13.09 1.96 0.022 3 - 6 38.55254 14.57 12.62 1.95 0.017 2 - 5 41.6
199
Table 6-3. Measured magnitudes, colors, and Brγ absorption and emission of potentialmassive stars in the embedded cluster around SGR 1627-41. Columns give StarID, instrumental K2.14-band magnitude, corrected Brγ - K2.14 (see §5.4.6), Brγ- K2.14 error, Equivalent Width, and distance from SGR 1900+14. Positivevalues for EW indicate absorption. Unless otherwise stated, errors in the K2.14-band magnitude are ± 0.01.
Star ID K2.14 (Brγ-K2.27)cal (Brγ-K2.27)cal error EW Distance(A) ′′
12456 20.18 0.05 0.02 7 - 14 37.112860 19.93 0.10 0.02 17 - 23 30.012981 20.15 0.10 0.02 16 - 22 20.013013 19.77 0.07 0.02 10 - 17 21.213079 16.03 -0.04 0.02 - (5 - 13) 32.613326 19.43 0.09 0.02 15 - 21 10.013813 20.62 0.07 0.02 9 - 17 34.113826 21.03 ± 0.02 0.08 0.03 10 - 20 19.113860 20.76 0.07 0.02 9 - 17 41.113894 19.11 0.08 0.01 13 - 18 30.413908 17.34 -0.03 0.02 - (2 - 9) 38.213951 19.27 0.07 0.01 11 - 16 9.413959 20.82 ± 0.02 0.08 0.03 9 - 21 36.814516 20.28 0.06 0.02 8 - 15 33.914665 20.23 0.07 0.02 10 - 17 33.814744 19.22 0.05 0.02 8 - 14 40.214844 18.79 0.05 0.01 7 - 12 36.814925 19.72 0.06 0.02 9 - 15 40.5
200
Table 6-4. Measured magnitudes, colors, and HeI absorption and emission for potentialOBI Stars in the embedded cluster around SGR 1627-41. Columns give StarID, instrumental K2.03 magnitude, corrected HeI - K2.03 (see §5.4.6), HeI - K2.03
error, Equivalent Width, and distance from SGR 1900+14. Positive values forEW indicate absorption. Unless otherwise stated, errors in the K2.03- bandmagnitude are ± 0.01
Star ID K2.03 (HeI −K2.03)cal (HeI −K2.03)cal error EW Distance(A) ′′
13959 21.12 -0.11 0.07 - (8 - 40) 36.813826 21.38 -0.09 0.05 - (8 - 32) 19.113813 20.97 -0.06 0.05 - (2 - 25) 34.113079 16.39 0.07 0.05 4 - 22 32.612504 17.64 0.12 0.05 14- 31 39.212705 16.64 0.17 0.04 23 - 39 33.012954 16.02 0.18 0.05 25 - 40 21.213801 15.96 0.17 0.05 23 - 39 44.214176 17.88 0.10 0.05 11 - 29 36.4
201
Figure 6-1. Histogram showing the number of stars at each calibrated magnitude for starsaround SGR 1900+14. A) Ks-band B) J-band. The bin sizes are 0.25 mag.Note the turnover at J ∼ 17 and Ks ∼ 14.5, indicating completeness (withinthe limits of crowding and confusion).
202
Figure 6-2. Color-color diagram of stars in the 4-band broad- and narrow- band catalogwithin a 5′ × 5′ subfield centered on SGR 1900+14. The vertical axis isflipped such that negative Brγ - Kcont values, which indicate Brγ emission,appear at the top half of the graph. Color denotes projected distance (in ′′)from SGR 1900+14, with purple indicating objects closer to the SGR andyellow indicating objects further away in 2-D space. The solid black line is thenarrow-band zeropoint (see §5.4.6) with m = 0.021 and b = -0.14. Bluer starson the left of the plot with more scatter in their narrow-band colors are at theedge of my detection limits and have more Poisson noise.
203
Figure 6-3. Brγ - K2.14 versus K2.14 color-magnitude diagram of the field surrounding SGR1627-41 The vertical axis is flipped such that negative Brγ - K2.14 values,which indicate Brγ emission, appear at the top half of the graph. Colordenotes projected distance (in ′′) from SGR 1627-41, with darker colorsindicating objects closer to the SGR and yellow indicating objects furtheraway in 2-D space. The solid black line is the narrow-band zeropoint (see§5.4.6) with Y = 0.18. Black circles are objects within a 45′′ radius of theSGR with Br γ absorption or emission. Triangles have both Br γ and HeIemission or absorption.
204
Figure 6-4. HeI - K2.03 versus K2.03 color-magnitude diagram of the field surrounding SGR1627-41. The vertical axis is flipped such that negative HeI - K2.03 values,which indicate HeI emission, appear at the top half of the graph. Colordenotes projected distance (in ′′) from SGR 1627-41, with darker colorsindicating objects closer to the SGR and yellow indicating objects furtheraway in 2-D space. The solid black line is the narrow-band zeropoint (see§5.4.6) with Y = 0.21. Black triangles are objects within a 45′′ radius of theSGR that have both have both Br γ and HeI emission or absorption.
205
Figure 6-5. Brγ - K2.14 versus K2.14 color-magnitude diagram of a field surrounding SGR0526-66 The vertical axis is flipped such that negative Brγ - K2.14 values,which indicate Brγ emission, appear at the top half of the graph. Colordenotes projected distance (in ′′) from SGR 0526-66, with darker colorsindicating objects closer to the SGR and yellow indicating objects furtheraway in 2-D space. The solid black line is the narrow-band zeropoint (see§5.4.6) with Y = 0.27. These data are within a 4′ search radius centered onthe cluster.
206
Figure 6-6. HeI - K2.03 versus K2.03 color-magnitude diagram for the field surroundingSGR 0526-66. On the vertical axis, positive HeI −K2.03 values, which indicateHeI absorption, appear on the top half of the graph. Color denotes projecteddistance (in ′′) from SGR 0526-66, with darker colors indicating objects closerto the SGR and yellow indicating objects further away in 2-D space. The solidblack line is the narrow-band zeropoint (see §§5.4.6) with Y = 0.22. Thesedata are within a 4′ search radius centered on the cluster.
207
Figure 6-7. Same plot as Fig 6-2 showing all 4-band catalog objects with Ks < 14.5magnitude. The candidate massive stars from Vrba et al. (2000) are markedwith inverted triangles.
208
Figure 6-8. A 1.5′ × 1.5′ region of our Ks-band image centered on Cl 1900+14. Previouslydiscovered OBI stars and newly identified massive stars detected by thissurvey are indicated – OBI stars from Vrba et al. (2000)are marked astriangles, new candidate massive stars as circles. The location of the SGR isalso noted.
209
Figure 6-9. Same plot as Figure 6-2 showing all 4-band catalog objects with Ks < 14.5magnitude. The candidate massive stars from Vrba et al. (2000) are markedwith inverted triangles. The new candidates from this study are marked withtriangles.
210
Figure 6-10. A 1.5′ × 1.5′ region of our Brγ image centered on Cl 1627-41. New candidatemassive stars exhibiting only Brγ emission or absorption are marked withcircles, stars exhibiting only HeI emission or absorption are marked by boxes,while sars with absoption or emission in both narrow-band filters areindicated by inverted triangles
211
Figure 6-11. Same plot as Figure 6-5. Stars within a 45′′ radius of the SGR are marked asboxes.
212
Figure 6-12. Same plot as Figure 6-4. Stars within a 45′′ radius of the cluster center ofSL463 are marked as circles. The lone OBI candidate is indicated with atriangle.
213
Figure 6-13. Same plot as Figure 6-5. Stars within a 45′′ radius of the SGR are marked asboxes.
214
Figure 6-14. Same plot as Figure 6-5. Stars within a 45′′ radius of the SL463 are markedas circles.
215
Figure 6-15. A 2.5′ × 2.5′ region of SGR 0526-66. The two circles indicate 45′′ search radiicentered on SGR 0526-66 and SL463, the potential natal cluster of themagnetar. The red box marks the location of the lone OBI star in thecluster, discovered by citetklo04.
216
CHAPTER 7LOCATIONS OF LUMINOUS BLUE VARIABLES IN THEIR HOST CLUSTERS
7.1 Introduction
Characterized as luminous, windy behemoths capable of mass loss rates upwards of
10−4–10−5 M¯ per year, Luminous Blue Variables (LBVs) are arguably some of the most
dramatically unstable and rare stars in the stellar zoo (Humphreys & Davidson 1994).
Recognized as evolved stars, LBVs are also thought to be at the extreme of the mass
spectrum; in fact one bona fide LBV (Eta Carinae) and two LBV candidates (the Pistol
Star and LBV 1806-20) vie for the title of most massive star in the Galaxy (Figer et al.
1998; Eikenberry et al. 2004a). As such, LBVs offer us a unique opportunity to study stars
near the theoretical upper mass limit and the compact objects that form when they die
(Eikenberry et al. 2004a).
Besides their distinguishing bolometric luminosities, large masses, and extreme mass
loss rates, LBVs are often identified by their emission-line rich spectra, photometric and
spectroscopic variability, and circumstellar ejecta (Humphreys & Davidson 1994). Of
these characteristics, stars that lack large amplitude variations (1-2 mag) but display
several other physical features are often identified as “candidate” LBVs, though the exact
segregation of candidates and confirmed LBVs remains somewhat subjective and may vary
between authors (Clark et al. 2005).
While LBVs have garnered a fair share of attention and study, the dearth of
information regarding their environments is notable. This is largely due to two factors
– the rarity of the stars themselves and the location of their environments. Found in the
dusty, crowded Galactic Plane of the Milky Way, the majority of massive star clusters
containing LBVs and other unusual massive stars have evaded detection by optical surveys
(Hanson 2003; Portegies Zwart et al. 2001). Only recently have advances in infrared
instrumentation enabled discoveries of elusive massive stellar environments.
217
The influx of new data include observations of clusters like Westerlund 1 and the
Quintuplet, posited to be the analogs of Super Star Clusters (SSCs) found in other
late-type galaxies, and Cl 1806-20, a smaller cousin to these objects (Clark & Negueruela
2002; Figer et al. 1999; Fuchs et al. 1999). All are notable for containing at least one LBV
among numerous Wolf-Rayets, supergiants, and hypergiants. Joined by the well-studied
Trumpler 16, these clusters offer us the first chance to identify characteristics of LBVs
found in cluster environments.
One particularly interesting property of LBVs in the aforementioned clusters is the
location of these massive stars with respect to other cluster members. During my study
of Cl 1806-20, I serendipitously noticed that LBV 1806-20 was located at the periphery of
the cluster. Given the mass of the LBV and the dictates of mass segregation (§7.4), one
would expect to find this object in the center. I checked several other LBV environments
and found upon visual inspection that four of five LBVs in young massive clusters appear
to be located on the periphery of their host clusters. The exception to this observation is
Eta Carina in Trumpler 16.
In this chapter, I evaluate the status of LBVs as “peripheral” cluster members, a term
I define is §7.3. I calculate the probability of finding at least one star in every cluster at or
outside the radius of the resident LBV and determine the robustness of my result using a
Monte Carlo simulation. Finally, I compare my finding to theoretical models that predict
the locations of massive stars in clusters and contrast these with scenarios that reproduce
the observed positions.
7.2 Sample Selection
For my study I require a homogeneous sample of objects: namely LBVs that are
established members of well-characterized clusters. While there are approximately 30
known Galactic LBVs, only five meet this criterion. The 15 known LBVs which have no
known cluster association were not included in my analyses. The rest were accepted or
rejected on a case-by-case basis, discussed below.
218
Four LBVs, Sher 25, ζ1 Sco, W51 LS1, and Cyg OB2 #12, are associated with large
star forming regions. Sher 25 is a member of the giant HII region NGC 3603, while W51
LS1 is associated with G49.5-0.4, a centrally located HII complex in W51 (Brandner et al.
1997; Okumura et al. 2000). ζ1 Sco and Cyg OB2 #12 are members of OB associations,
Sco OB1 and Cyg OB2, respectively (van Genderen et al. 1984; Hanson 2003).
Both NGC 3603 and Sco OB1 are hosts to young cluster of massive stars; the dense
starburst cluster NGC 3603 YC is centrally located in NGC 3603 and the open cluster
NGC 6231 resides in ScoOB1 (Stolte et al. 2004; van Genderen et al. 1984). While the
membership of ζ1 Sco in Sco OB1 and Sher 25 in NGC 3603 is well established, neither
LBV’s membership in the resident cluster is certain; mostly because the LBV is a notable
distance from the cluster center. Without conclusive evidence to establish that the LBVs
are de facto cluster members, I excluded these LBVs from my sample.
While Cyg OB2 #12 is a confirmed member of the famed Cyg OB2 (Hanson 2003),
the lack of a young cluster within the OB association prevents us from including the LBV
in my study of cluster members. W51 LS1 was excluded for the same reason. While home
to many massive OB stars, including W51 LS1, the G49.5-0.4 complex in W51 has no
known central cluster (Okumura et al. 2000). While in both instances I observe that the
resident LBV is toward the edge, rather than the middle, of its host region, I recognize
these locales are significantly different from the cluster environment I sought to explore.
Several LBVs and candidate LBVs are members of the Galactic Center Cluster.
This extreme environment is highly unique; centered around a super-massive black hole
and subject to strong tidal forces, magnetic fields, and stellar winds, the cluster’s star
formation history is far from understood. For instance, whether the cluster formed in situ
or spiraled inward is still a matter of debate (Paumard et al. 2001). Clearly, the positions
of the LBVs might result from any number of mechanisms unique to the Galactic Center
region. Therefore the Galactic Center Cluster is not comparable to the homogenous
clusters I assess here.
219
Four LBVs, HD 80077, P Cygni, and WRA 751, are associated with massive stars
or star clusters in their vicinity. Moffat & Fitzgerald (1977) identified HD 80077 as a
potential member of Pismis 11, noting that several suspected cluster members were
reddened by the same amount as the LBV. However, since both the constituency of Pismis
11 and the membership of the LBV remain unclear I exclude it from my sample.
The status of P Cygni is akin to that of HD 80077. Turner et al. (2001) claim that
P Cygni is a member of an unnamed grouping of stars that form a double cluster with
the young open cluster IC 4996. P Cygni’s membership in the small unnamed grouping
is inferred from line-of-sight and reddening arguments, similar to those made in the case
of HD 80077. Since little is known about the stars in the unnamed cluster and cluster
membership of P Cygni is not confirmed, I did not include it in my sample.
WRA 751 appears to be a member of an unnamed cluster discovered by Pasquali
et al. (2006). Again, similarities in reddenings and line-of-sight arguments are used to
identify the LBV as a member of the cluster. However, this potential cluster is far from
well-established. The authors point out that several stars distant from the cluster core
have reddenings similar to the cluster proper, but may not be members. Furthermore,
the techniques used to cull non-cluster objects also excludes fainter stars, even if they are
indeed cluster members. Given the paucity of data about this newly discovered cluster
and the need for further observation to establish the LBVs membership and the cluster
census, WRA 751 was rejected from my sample.
AG Car is located in a particularly dense field of massive stars; between 50-100
massive stars are located within 20′′ of the LBV (Hoekzema et al. 1992). Extensive studies
by Hoekzema et al. (1992) found that while AG Car lies in the line-of-sight of both Car
OB1 and Car OB2, it is a member of neither association. In fact, while they identified
a few potential WR stars at a distance consistent with AG Car, they found no massive
cluster associated with the LBV.
220
Thus, five Galactic LBV meet my criteria as confirmed members of well-characterized
clusters. I present a summary of relevant characteristics and literature sources for each
cluster below:
Located between 2-4 kpc away, Westerlund 1 spans 0.6-2.6 pc and has ∼ 60 massive
members. Counted in this census is W243, a candidate LBV (Clark & Negueruela 2002).
I note that W243 is outside the 50′′ diameter core defined in a recent survey of the cluster
(Clark et al. 2005). In this letter, I use the positions of massive cluster members from
several recent studies of Westerlund 1 (Clark et al. 2005; Negueruela & Clark 2005;
Crowther et al. 2006). Cluster membership is well-defined; Clark et al. (2005) establish
the massive cluster population with optical photometry and spectroscopy, while Crowther
et al. (2006) confirm the spectral type of cluster Wolf-Rayets with infrared narrow-band
imaging and spectroscopy.
The Quintuplet, known for the five bright stars that lend the cluster its name,
contains two potentially peripheral LBVs, the Pistol Star, and FMM 362 (Figer et al.
1995, 1999). At the distance of the Galactic Center, the cluster spans 50′′ and is composed
of > 30 massive stars and potentially hundreds of unseen low mass stars. The Pistol Star
and FMM 362 are located approximately 30′′ and 45′′ from the core. I use positions of
the brightest members of the Quintuplet from Figer et al. (1995), who identified cluster
members based on near-infrared photometry and K-band spectroscopy.
Another recently discovered massive cluster, Cl 1806-20, contains LBV 1806-20
(Eikenberry et al. 2004a; Figer et al. 1999). Smaller than the Quintuplet or Westerlund 1,
it hosts ∼ ten massive stars (LaVine et al. 2003; Figer et al. 2005). Located at a distance
of ∼ 15 kpc, it spans less than 40′′ on the sky. A cursory view of the distribution of cluster
stars reveals that the LBV is 30′′ from the densest region of Cl 1806-20. I utilize positions
of massive stars from Figer et al. (2005), who confirmed the membership of objects in Cl
1806-20 using near-infrared spectroscopy.
221
Regarded as the center of the open cluster Trumpler 16, the well-studied LBV, Eta
Carina, provides a contrast to the rest of the sample. At a distance of 3.6 − 4.0 kpc,
Trumpler 16 spans greater than 5′ and is home to ∼ 30 massive stars and hundreds of
lower mass members (DeGioia-Eastwood et al. 2001; Carraro et al. 2004). Using positions
of Trumpler 16 cluster members from DeGioia-Eastwood et al. (2001) I complete the
analyses described in §7.3.1 and §7.3.2.
I present images of each cluster, indicating their resident LBV, in Figure 7-1.
7.3 Probability Analyses
While a cursory visual inspection indicates that Cl 1806-20, W243, FMM362, and the
Pistol Star are not located in the cores of their clusters but the outskirts (see Figure 7-1),
I sought to verify this qualitative impression with quantitative evidence. My first task was
to define what it means to be a peripheral cluster member.
Mathematically assessing my initial visual observation on a cluster by cluster basis
proved to be challenging. For example, I cannot use a 3σ deviation from the mean
radius to define “peripheral”, as no cluster members in the literature data meet this
criterion. Any object located this far away from the cluster center would most likely not
be considered a cluster member. Furthermore, small number statistics create additional
problems. The clusters in my sample have few stars; assessing the statistical significant
of the position of one star out of ten is difficult. Although this is the case, I recognized
that it is difficult to explore the potential of LBVs as peripheral cluster members without
addressing where LBVs lie in each host cluster and how their locations compare to other
cluster stars. With this in mind, I sought to carefully characterize the positions of LBVs
within individual clusters (§7.3.1).
While this analysis cannot confirm whether individual LBVs are peripheral cluster
members to a high statistical certainly, it motivated the further exploration of LBVs as
an ensemble of peripheral objects. In the context of a statistically significant sample,
an ensemble of objects are peripheral cluster members if they have a broader radial
222
distribution about the mean than similarly massive stars in their clusters. I explore LBVs
as an ensemble of peripheral objects in §7.3.2.
7.3.1 Individual Analyses of LBVs as Peripheral Cluster Members
Mass segregation theories suggest that less massive stars should move toward the
cluster outskirts while massive stars sink toward the center (Spitzer 1987; Binney &
Tremaine 1987). I seek to answer a specific question; are Luminous Blue Variables
anomalously peripheral with respect to mass segregation theories? Judging whether or not
an object is in an anomalous location requires a comparison population. In this case, I
compare the resident LBV to the remaining massive stellar population in their respective
host clusters. Are LBVs further from the center than the majority of the massive stars?
Furthermore, does the area enclosed within the radius at which the LBV is located contain
the bulk of the mass associated with the massive population?
Using the literature data described in §7.2, I found the average right ascension and
declination of each cluster using the position of its members; this gave us a de facto
cluster center. I then found the distance between each cluster star and this central
position. Finally, I counted the total number of stars with a distance greater than or equal
to that of the LBV. I divided this number by the total number of cluster stars and called
the resulting value ζLBV . I list my values for ζLBV in Table 7-1. In each cluster more than
75% of the other massive stars in each cluster are enclosed within the radius of the LBV.
I then took this analysis a step further. Using a recent paper by Crowther (2007)
I estimated masses for Wolf-Rayet stars in each LBV host cluster. The author reported
that WC stars spanned a mass range of 9 − 16M¯ while WN stars spanned 10 − 83M¯.
I assigned the average value of 12.5M¯ to every WC and 46.5M¯ to each WN. I then
assigned masses to all cluster supergiants using values from Cox (2000). I recomputed
the position of the cluster center of each cluster, weighting the position of every star by
its stellar mass. I found the distance to the LBV from the mass-weighted cluster center
and calculated the percentage of the total mass associated with massive stars which is
223
enclosed by the radius of the LBV. Finally, to eliminate any potential bias resulting from
the assumption of Crowther (2007) masses for WC and WN masses, I repeated the above
analysis using uniform values for both types of Wolf-Rayets. In the second iteration I
used the average mass of WR stars, 46.5M¯ for every Wolf-Rayet. In the final case, I
used a value on the low end of the mass range, 25M¯. In all calculations, the total cluster
mass associated with massive stars excludes all LBVs in the cluster. I present the results
in Table 7-2. I noticed that while for three LBVs in my sample, the percentage of mass
enclosed within the LBV radius is fairly unaffected by my assumptions for WR masses,
the values for one LBV, the Pistol Star, are more sensitive. However, in all iterations, the
majority of the total cluster mass associated with massive stars is always enclosed by the
LBV radius.
7.3.2 Ensemble Analysis of LBVs as Peripheral Cluster Members
In the previous section, I established that LBV 1806-20, the Pistol Star, FMM 362,
and W243 do indeed lie outside the majority of massive cluster members and the bulk
of the cluster mass associated with massive stars. I point out that if these stars were not
members of a particular stellar type, my result would not be particularly interesting.
Several stars must be at or near the edge of a cluster and one star must be furthest away;
clearly, finding a star on the outskirts of a cluster is, in itself, unremarkable.
What makes my results intriguing is the context. It seems unlikely that four stars of
the same rare class, would, by chance alone, lie in the outskirts of their cluster. However,
I wish to analyze this “hunch” more methodically. Thus, I calculated the probability of
consistently finding five stars, one in each cluster, outside the radius of the resident LBV.
I define this value as ζtotal. A low probability, a small value of ζtotal, would imply that the
positions of the LBVs was not merely random.
To determine the total probability, ζtotal, I first measured the likelihood of finding
a star at or beyond the cluster radius of each individual LBV. This probability, ζLBV , is
simply the number of stars lying outside the radius of the LBV and was calculated for
224
my analysis in §7.3.1. The results appear in Table 7-1. The product of the five values of
ζLBV is equal to ζtotal. In all cases, my number counts exclude all cluster LBVs. When
calculating values ζLBV for the two LBVs in the Quintuplet, I used the radius of the Pistol
star for ζPistol and the radius of FMM 362 for ζ362.
Using my values of ζLBV , I gauged the likelihood of finding at least one star in each
cluster outside the radius of LBV, the goal of this analysis. I multiplied the five values of
ζLBV in Table 7-1 and found ζtotal = 1.9× 10−4. This value implies that there are only two
chances in ten thousand that when randomly selecting n massive stars from n clusters, all
will be at or beyond the radius of the LBV in each cluster. My value of ζtotal suggests that
it would be difficult for random chance alone to account for the positions of the LBVs.
However, before I accept this conclusion, I must address an important detail. ζtotal is
the product of five values of ζLBV ; can drawing five numbers at random from a uniform
distribution and multiplying them together yield a number less than or equal to 1.9 ×10−04? If I find that the product of five random numbers between zero and one easily
reproduce probabilities of 0.02% or 0.002%, my result would be less robust.
I evaluated this hypothesis using a Monte Carlo simulation. I choose a number
at random from a uniform distribution for each cluster. This simulated value of ζLBV
is defined as ζsim−LBV . This number represents an object’s rank, in distance, from
the center. I choose a uniform distribution so that my analysis is independent of the
functional form of the spatial distribution of stars in the cluster.
Multiplying five values of ζsim−LBV , I computed a random value for ζsim−total, my
simulated value of ζtotal. I repeated this process for ten thousand iterations, resulting
in ten thousand values of ζsim−total. I then calculated the number of times I found a
value less than or equal to ζtotal by random chance alone and divided by my number of
iterations. I found that only 7% of the ten thousand simulated values of ζsim−total were less
than or equal to the value of ζtotal derived from the observations. Thus, my Monte Carlo
225
simulation indicates that my observed value of ζLBV is difficult to reproduce in a random
distribution.
7.4 Discussion
The techniques described in the previous section confirm my initial observations, that
the majority of LBVs in well defined young clusters are located on cluster outskirts. I
found that LBVs, as an ensemble, met the definition of peripheral cluster member defined
in §7.3.2, even when including the control cluster Trumpler 16. I calculate a two in ten
thousand chance that n randomly selected massive stars from n clusters will all be at or
beyond the radius of the LBV in each cluster. I find that 7% of the time I can obtain a
value less than or equal to my value for ζtotal by random chance, which implies that my
result is at approximately a 2σ level.
Having established that random chance alone can not account for the observed
position of LBVs in clusters, I must seek physical explanations. One might ask, where
do I expect to observe LBVs in clusters? Mass segregation theories posit that massive
stars in clusters sink into the center while less massive stars are relegated to the outskirts
of the cluster (Spitzer 1987; Binney & Tremaine 1987). Furthermore, the massive stars
that sink to the center are likely to suffer collisions, form binaries, and participate in
runaway mergers. Thus, the most massive stars, often the product of multiple collisions,
are expected to be centrally located (Portegies Zwart & McMillan 2002; Portegies Zwart
et al. 1999).
I recall that the LBVs are generally regarded as the most massive stars in the Galaxy.
The five LBVs in my analysis are easily amongst the most massive in their respective
clusters, with reported masses > 100M¯ (Figer et al. 1998; Eikenberry et al. 2004a; Clark
& Negueruela 2004). Clearly, given the extreme masses of LBVs and the dictates of mass
segregation theories, one would suppose that LBVs would be centrally located. To place
my result in context; theory suggest that LBVs should not be on the outskirts of their
cluster. Mass segregation strongly implies that for LBVs, ζtotal should approximate one;
226
almost every star should lie further than the LBV from the center. Furthermore this
should be at least a 3σ result. Thus my result is strengthened; it lies in the opposite
direction of what theory predicts.
One significant exception to my result is Eta Carina. which appears centrally located
in its cluster. However, I note that since I cannot view the three-dimensional structure
of Trumpler 16, it is possible that line-of-sight projection effects could account for Eta
Carina’s apparent central location. In the absence of dynamical evidence I can not further
comment on this possibility, but it is important to recognize that I also cannot guarantee
that Eta Carina is actually in the center of Trumpler 16.
My results pose several questions. What scenarios might lead to LBVs being located
in the periphery rather than the core of the cluster? Is one particular process at work in
these massive clusters or could several factors operate to produce these notable exceptions
to the mass segregation rule?
While the clusters themselves have similar taxonomy, the Galactocentric distances
vary widely. The Quintuplet, for instance is located 35 pc from the the Galactic Center
(Portegies Zwart et al. 2001). Might tidal forces influence the location of the LBVs in this
cluster? Portegies Zwart et al. (2001) noted that tidal effects have considerable effects
that increase as Galactocentric distance decreases. Indeed, the Quintuplet is believed
to be just barely bound against tidal disruption (Figer et al. 1999). While this is not a
conclusive answer, it suggests that other forces might deter the normal infall of several of
this cluster’s massive star.
However, neither Westerlund 1 nor Cl 1806-20 is within a Galactocentric distance
that might suffer from these tidal effects. How can I explain the positions of the LBVs
in these clusters? One point of interest is that both clusters contain magnetars: highly
magnetized neutron stars. Recently, several authors have suggested that the progenitors
of these objects may be very massive stars (Eikenberry et al. 2004a; Fruchter et al. 2006).
If future works confirm that very massive stars are the antecedents of magnetars, their
227
presence in Westerlund 1 and Cl 1806-20 would mean that the current LBVs in these
clusters were not always the most massive cluster members.
In this scenario, the current LBV would not dominate the dynamical evolution of its
cluster and therefore would not necessarily be the first star in the core. Still, even as the
second or third most massive star in the cluster, one would still expect the LBV to sink
to the center (Portegies Zwart & McMillan 2002). However, the presence of another, more
massive, potentially violently-evolving star sinking toward the core offers the potential for
interactions and ejections. This scenario has an important caveat; initial observations of
the clusters indicate that that the magnetar in Westerlund 1 is greater than 1′ from the
cluster core and also, that SGR 1806-20 is not centrally located (Kaplan et al. 2002; Figer
et al. 2005; Muno et al. 2006). It is still possible, though, that the presence of evolved
and/or supermassive stars in Westerlund 1 and Cl 1806-20 during a previous epoch could
offer some explanation for LBV positions.
Alternatively, my result might offer a glimpse into ongoing theoretical discussions
regarding primordial mass segregation in clusters (Bonnell et al. 2001; Bonnell & Bate
2006). One important tenet of this theory is that massive stars are more likely to form
in the cores of star forming regions, suggesting that even in very young clusters, massive
stars should be centrally located. This prediction is at odds with the observed peripheral
location of four of the massive LBVs in my sample. Although at this point I can draw no
definite conclusions, I point out that these observations of the locations of massive stars
in 104 − 105M¯ mass clusters may help refine theoretical models of cluster evolution and
primordial mass segregation.
Finally, I address the LBVs rejected from my sample. Several authors suggest
that many of the massive clusters in our Galaxy remain undiscovered (Hanson 2003;
Portegies Zwart et al. 2001). It is certainly possible that LBVs with no current cluster
associations may, in fact, be members of undetected clusters. If new clusters are
discovered around these LBVs, it would be useful to extend my study to include them.
228
Additionally, confirming the host clusters of HD 80077, AG Car, P Cygni, and WRA 75
and characterizing the full extent of the cluster’s massive star population would allow
us to expand my study to include them, increasing my sample. Finally, if I establish
cluster membership of Sher 25 or ζ Sco, two LBVs that may be peripherally located in
young clusters, adding these LBVs to my analyses may significantly strengthen my results.
Clearly a good deal more data are needed before I can say with certainty that LBVs
in clusters are preferentially peripherally located or explain why those I discuss in this
chapter are exceptions to the rule.
7.5 Conclusion
My analyses show that LBVs in Westerlund 1, Cl 1806-20, and the Quintuplet are
located on the periphery of their host clusters. I also show that these LBVs are not likely
to be located on the outskirts of their clusters by chance alone; only a two in ten thousand
chance exists that that n randomly selected massive stars from n clusters will all be at or
beyond the radius of the LBV in each cluster. Furthermore I argue that the locations of
these LBVs is contrary to theoretical expectations, as mass segregation theories suggest
that massive stars should quickly move to the cores of their natal clusters.
I offer potential explanations for the locations of the cluster LBVs, but conclude that
the current lack of data currently available offers no answers or single mechanism to easily
account for their positions. Increasing the sample size by exploring the environments
around seemingly isolated LBVs, establishing LBV membership in potential host clusters,
and expanding our knowledge of LBVs in new or poorly studied environments may offer us
the best chance to determine if LBVs are indeed preferentially located on the periphery of
massive clusters.
229
Table 7-1. Position of LBVs in clusters compared to other cluster members. Columnsindicate Luminous Blue Variable name, host cluster name, total number ofmassive stars in host cluster excluding all cluster LBVs (NMS), radius of LBVs(RLBV ) in arcseconds and pc, number of stars outside the LBV radius(NR>LBV ) excluding all cluster LBVs, and number of stars outside the LBVradius divided by total number of massive stars in each cluster (ζLBV ). Iassumed a ζLBV = 1 for our control LBV, Eta Carina in Trumpler 16.
LBV Cluster NMS RLBV RLBV NR>LBV ζLBV
(′′) (pc)1806-20 Cl 1806-20 10 17.0 1.24 1 0.100W243 Westerlund 1 62 94.7 2.5 6 0.097Pistol Star Quintuplet 35 31.1 1.2 8 0.229FMM362 Quintuplet 35 45.0 1.7 3 0.086
Table 7-2. Percentage of cluster mass associated with massive stars enclosed bymean-weighted LBV radius. Columns indicate Luminous Blue Variable name,host cluster, mass-weighted radius of the cluster LBV (RMW−LBV ), andpercentage of cluster mass associated with massive stars enclosed by the LBVmass-weighted radius, % Menc, for each of the three trials.
Crowther (2006) WR = 46.5M¯ WR = 25M¯LBV Cluster RLBV−MW % Menc RLBV−MW % Menc RLBV−MW % Menc
1806-20 Cl1806-20 17.2 91 18.4 93 18.7 90W243 Westerlund 1 98.2 91 94.8 88 96.9 91Pistol Star Quintuplet 33.5 86 30.3 72 31.4 77FMM 362 Quintuplet 45.9 91 47.7 94 46.9 95
230
Figure 7-1. Clockwise starting from the upper left. Images of Westerlund 1 (Clark et al.2005), the Quintuplet (Figer et al. 1999), & Cl 1806-20. In each image theLBV is circled and labeled and the cluster center is marked.
231
CHAPTER 8CONCLUSION
8.1 Summary and Motivation of the CIRCE Project
Since the invention and refinement of HgCdTe detectors and infrared optimized
telescopes, near-infrared instruments have become a mainstay of astronomical research.
Currently, ten of the world’s largest telescopes have a total of ∼ 20 near-infrared
instruments available for observation or in the final stages of integration. Plans for
bigger and better instruments, including the next generation of NIR cameras designed
for 20 and 30 meter telescopes, are already underway (Eikenberry et al. 2006a). With
so many NIR instruments currently in operation and more on the horizon, what is the
justification for building CIRCE?
As discussed in Chapters 1 and 2, CIRCE’s first function is to fill the gap between
the inception of science operations at the GTC and the commissioning of the facility
NIR instrument EMIR. During that time, CIRCE will be the only NIR instrument
on the world’s largest optical/infrared telescope. Anticipating that it will be used
for a wide-variety of science projects, we designed CIRCE as a workhorse instrument
with multiple modes. Once EMIR is complete and ready for use on the GTC, CIRCE
will move on to its secondary role – to complement the suite of facility instruments
with spectro-polarimetric modes, high-speed photometry, and low- and mid- resolution
spectroscopy.
Besides these drivers, CIRCE has a third purpose. CIRCE was conceptualized as a
training instrument for students pursuing careers as instrument scientists, astronomers
who guide the development of instruments to ensure they meet the scientific needs of
the astronomical community. While new instruments often explore the “bleeding edge”
of technology, they are fundamentally driven by the science needs of the astronomical
community. While in theory any number of designs might be feasible for a given
instrument, instrument scientists advocate solutions that will produce the best data. This
232
requires not only a working knowledge of a wide-variety of astronomical projects and the
type of observations necessary to carry-out each experiment, but the ability to translate
these requirements into optical, mechanical, and electronic designs and specifications.
However, at a time when the horizon of astronomical instrumentation is rapidly
expanding and new instrument scientists are needed, the liability, complexity, and expense
associated with facility instruments has significantly limited student participation in these
projects. One of the basic tenants of CIRCE was to allow graduate students critical roles
in the design and integration of an instrument destined for an 8 - 10 meter class telescope.
In Chapters 2, 3, and 4, I detailed my contributions to the optical, opto-mechanical,
and cryo-mechanical design of CIRCE. In Chapter 2, I presented the development
of CIRCE’s all-reflective aspheric optical design, specifically my role in the redesign,
analysis and tolerancing of the system (Edwards et al. 2004). In Chapter 3, I introduced
the concepts behind mechanical design and discussed my creation of two- and three-
dimensional drawings of mirrors, benches, and brackets (Edwards et al. 2006). Finally,
I outlined my work on the solid models and two- dimensional drawings of the pupil box
mechanism in Chapter 4 (Edwards et al. 2006).
CIRCE, like essentally all astronomical instrumentation, is a team effort. Two
other students, Miguel Charcos-Llorens and Nestor Lasso, and a postdoctoral fellow, Dr.
Antonio Marin-Franch each worked on a wide variety of other sub-systems and modes.
Dr. Marin-Franch designed the software and user control interface that operate cryogenic
motors, temperature sensors, and other mechanisms inside the CIRCE dewar. Nestor
Lasso led the electrical layout and helped with the modification of the detector array.
Miguel Charcos-Llorens developed the polarimetric mode, created a program to measure
the flexure of the bench, and designed the cryostat and focal plane mechanism. He also
completed the optical bench and modified the pupil box to meet new specifications
required by the polarimetric design. I discuss the current status in more detail in §8.4.
233
8.2 Status of CIRCE
The CIRCE team has completed the design of many critical CIRCE opto- and cryo-
mechanical components including the pupil box and focal plane mechanisms, brackets,
mirrors, cryostat, and detector stage. Most of the manufacture will take place in the UF
Physics and Astronomy Machine Shop, including the optical bench. To substantially
reduce cost and lead time, M. Charcos designed the large (1 meter diameter × 1.5 meter
long) optical bench in multiple pieces so that it could be manufactured on machines
available in-house at UF. Flexure modeling shows that this option is viable; the large
cryogen tank beneath the bench offers adequate stiffness to support the optics and
cryo-mechanisms. We anticipate that completion of the bench and mirror blanks in the
next three months and the remaining components approximately three months after.
Once the bench and mirror blanks are complete, we will ship them to Janos
Technology, who will diamond-turn the mirrors on a CNC lathe and mount them on the
brackets and bench. They will then employ a newly developed technique that emphasizes
ensemble testing of the mirrors, brackets, and bench. After testing for errors, they will
slightly reshape the mirrors to reduce aberrations and improve image quality. This novel
solution to optical analysis allows for immediate correction of alignment problems in situ.
Thus the completed bench, with mirrors and brackets in place and aligned, will arrive at
UF ready for integration. We anticipate that the manufacture and alignment of the optics
will take 3 - 6 months.
The design of the electrical layout is in the final stages of review. Cryogenic motors
and the hardware to control them are currently in-hand at UF. Control software, along
with a Java-based User Interface, designed by A. Marin-Franch, is complete. The interface
was rigorously tested by the IAG, who used it extensively during the integration of
Flamingos 2. Testing of cryo-mechanisms, creation of cables, and manufacture of the
cryostat by an off-site vendor will take place in the next 6 months during the manufacture
and alignment of the optics off-site.
234
One all of CIRCE’s components return to UF, we anticipate a further 3 - 6 month
integration and testing period. Thus, we expect that CIRCE will ship to the telescope
sometime during the late summer or early fall of 2009.
8.3 Summary of A NIR Narrow-Band Imaging Survey
Utilizing the ability of NIR radiation to penetrate highly extinguished environments,
I completed a narrow-band survey to search for the emission lines which are signatures
of massive stars in the field surrounding Soft Gamma Repeaters (SGRs). By probing for
clusters that host both massive stars and magnetars, I sought to link the births of these
rare neutron stars with the deaths of very massive stars.
Using the current generation of NIR instruments, including the ISPI on the CTIO 4
meter and WIRC on the Palomar 200”, I observed the fields around SGR 1806-20, SGR
1900+14, SGR 1627-41, and SGR 0526-66. To test whether my NIR narrow-band survey
would successfully find massive star candidates, I first applied my technique to SGR
1806-20, a member of the known massive star cluster, Cl 1806-20 (Chapter 5; Edwards et
al. submitted). My method, which utilized PSF photometry of broad- and narrow- band
images to create a color-color diagram, successfully picked out a number of known WR
and OBI stars. It also identified ∼ 10 OBI candidates for follow-up spectroscopy.
In Chapter 6, I apply the technique developed in Chapter 5 on the remaining three
SGR fields. First I studied the embedded cluster associated with SGR 1900+14. Two
potential cluster members, both M5 supergiants, saturated the central region of the frame,
rendering it difficult to study previously identified cluster stars or find new massive cluster
members in the area. However, I identified > 10 potential OBI and rare emission-line
stars within 45′′ of the SGR (Edwards et al. in preparation). Next, I studied the field
surrounding SGR 1627-41, a virtually unexplored magnetar environment. I found a
number of potential massive stars with Brγ absorption and emission. Several of these also
exhibit HeI features, making them strong candidates for follow-up spectroscopy.
235
Finally, I studied two regions around SGR 0526-66: a 45′′ search radius centered on
the magnetar and a 45′′ search radius centered on SL 463, a nearby massive star cluster.
While I did detect one bright candidate OBI star in the center of SL 463, confirming the
observations of Klose et al. (2004), I did not find a population of stars with narrow-band
emission or absorption in either region. However, obscurration by local dust, especially
given the existence of SNR N49 near the magnetar, could easily skew my photometry.
Deeper narrow-band observations and the addition of broad-band data is clearly necessary
before drawing conclusions about the stellar populations near SGR 0526-66.
While narrow-band imaging is sometimes suggested as potentiall simpler and faster
way to search for massive stars without performing spectroscopy, my work shows that
this technique requires extensive calibration, excellent quality narrow- and broad-
bandphotometry, and careful analysis to yield reliable and unambigious results. However,
once these requirements are met, the technique is viable and potentially powerful,
especially in very crowded environments where color-color and color-magnitude diagrams
can reduce the number of massive cluster candidate from thousands to tens.
8.4 Future Work
8.4.1 Extinction and Stellar Populations in the SGR Fields
One bi-product of my search for massive stars around SGRs is excellent broad-band
photometry of thousands of stars in the regions surrounding SGR 1900+14 and SGR
1806-20. I plan to use these data to study variations in the extinction law along the
line-of-sites to the clusters. This might be especially enlightening in the case of SGR
1900+14, where controversy concerning the reddening to the cluster exists. Further, I will
examine populations of stars at different reddenings to construct more complete pictures
of the stellar populations of the fields. Finally, I will search for unrelated serrendipitous
clusters by looking for overdensities in the field.
236
8.4.2 Massive Stars with CIRCE
While I completed the bulk of my narrow-band imaging survey of SGR environments
using the previous generation of NIR instruments, I anticipate that observations
with CIRCE on the GTC will provide the keystone for this project. Using CIRCE’s
spectroscopic mode, I will perform long-slit grism spectroscopy to follow-up all massive
star candidates. Comparing this spectroscopy to NIR atlases of massive stars (Hanson
et al. 1996; Figer et al. 1998), I will determine the spectral type of each identified
massive stars and estimate its distance and reddening, confirming each candidate’s
cluster membership.
I will also use CIRCE to further investigate the fields surrounding LBVs. In Chapter
7, I showed that LBVs in Westerlund 1, Cl 1806-20, and the Quintuplet are located on the
periphery of their host clusters. I presented statistical evidence to demonstrate that these
massive stars are not likely to be located on the outskirts of their clusters by chance alone.
However, given the lack of LBVs with firmly established cluster membership, the need
for more observations is clear. Using CIRCE narrow-band imaging and the techniques
demonstrated in this dissertation, I will probe for massive stars around seemingly isolated
LBVs or those in poorly studied regions. I hope to increase the population of LBVs
associated with massive clusters and then reassess the locations of these objects within
their host clusters using a more statistically significant sample.
8.4.3 Magnetar Environments
There are two types of magnetars; Soft Gamma Repeaters and Anomalous X-ray
Pulsars. In this dissertation, I focused on SGRs, which have the most evidence for
potential associations with massive clusters. Located in regions of heavy extinction, their
environments offered the best opportunity to use NIR observations to find large massive
clusters previously missed by optical surveys.
Similarly, many AXPs are also in the Galactic Plane where extinction is patchy
and crowding problematic. Thus next logical step is to expand my probe of mangetar
237
environment to include AXPs. If I find massive stars around AXPs, it will strengthen the
argument that magnetars form from very massive stars that lost most of their mass during
their LBV and WR phases. If I confirm that all SGRs are members of massive clusters,
but fail to find massive stars around most AXPs, this may suggest different formation
mechanisms for the two “flavors” of magnetars. Thus, investigating AXP environments
should provide substantial information about the formation channels of magnetars and the
differences between SGRs and AXPs.
I have started an observing campaign to survey AXP environments with narrow-band
imaging using currently available NIR instruments. To date, I have observed four AXP
fields with WIRC using Brγ and 2.27 µm Kcont filters. In summary, I plan to continue
this survey, expand it to include data from CIRCE on the GTC, perform follow-up single
slit and multi-object spectroscopy of both SGR and AXP massive cluster candidates, and
begin a new narrow-band survey to survey the environments of LBV. I believe that these
projects will provide significant results on a variety of enigmatic objects.
238
APPENDIX: ZEMAX OUTPUT OF CIRCE PRESCRIPTION
The CIRCE optical prescription consists of several tables that describe the locations,
rotations, and general properties of each surface. The most important of these is the
Surface Data Summary, which I show in Table A-1. This table gives a very detailed
description of every mirror in the prescription; listing values of the radius, thickness, glass,
diameter and conic constant. Conic and flat mirrors are listed as “standard” surfaces
while the higher order asphere is listed as “evenasph”.
Coordinate breaks (“coodbrk”) are also listed as a type of surface in the Surface
Data Summary. In the Lens Data Editor, shown in Figure 2-4, columns for the X, Y, Z
decenters and α and β tilts are listed for all coordinate breaks that appear in Table A-1.
However, in the optical prescriptions these values are output into a different table, the
Coordinate Break Summary, which I show in Table A-2.
Three surfaces always appear in the Lens Data Editor and Surface Data Summary:
OBJECT, STOP, and IMAGE. The OBJECT is the surface from which ZEMAX launchs
the rays that will propogate into the system. The OBJECT is always surface 0. The
STOP is the aperture stop of the system. In most IR telescopes, the secondary mirror acts
is defined as the aperture stop. Examining Table A-1, one can see that the Surface labeled
STOP is listed as the secondary. The IMAGE is where the image of the object is formed;
this is where ZEMAX performs most of the image quality diagnostics
The prescription can also include other tables, like the field angles and wavelengths
used to perform optical analysis. While these are important in determining the quality of
the instrument, they do not impact that layout of the camera.
239
Table A-1. ZEMAX Surface Data Summary
Surface Type Comment Radius Thickness Glass Diameter ConicOBJ STANDARD Infinity Infinity 0 01 STANDARD Infinity 0 10159.09 02 STANDARD Infinity 14740 1090 03 STANDARD PRIMARY -33000 -14739.41 MIRROR 10144.61 -1.00225STO STANDARD SECONDARY -3899.678 18070.61 MIRROR 1070 -1.5048355 STANDARD Infinity 20 FSILICA 241.2386 06 STANDARD Infinity 55 240.6069 07 STANDARD FOCAL PLANE Infinity 130 238.6015 08 COORDBRK D1 (A) - 0 - -9 STANDARD F1 Infinity 0 MIRROR 289.4733 010 COORDBRK D1 BEND (A) - -279 - -11 COORDBRK D2 (A) - 0 - -12 STANDARD F2 Infinity 0 MIRROR 313.5647 013 COORDBRK D2 BEND (A) - 790.9824 - -14 COORDBRK D3 (Y) - 0 - -15 COORDBRK D3 (A) - 0 - -16 STANDARD C3 -1466.211 -534.2299 MIRROR 831.7695 -0.33542317 STANDARD C4 -2771.235 251.5313 MIRROR 300.0933 -47.7546518 COORDBRK D4 (Y) - 0 - -19 COORDBRK D4 (A) - 0 - -20 COORDBRK D5 (Y) - 0 - -21 STANDARD D5 (Z) Infinity 5 54.55716 022 COORDBRK D5 (A) - 0 - -23 STANDARD LYOT Infinity 0 53.35195 024 COORDBRK D5 RET (-A) - 0 - -25 STANDARD D5 RET (-Z) Infinity -5 53.37644 026 COORDBRK D5 RET (-Y) - 0 - -27 STANDARD FILTER Infinity 3 FSILICA 54.94246 028 STANDARD Infinity 272 54.39428 029 COORDBRK D6 (G) - 0 - -30 STANDARD Infinity 368 128.4977 031 COORDBRK D7 (Y) - 0 - -32 COORDBRK D7 (A) - 0 - -33 STANDARD IM5 -453.7226 -208.8597 MIRROR 625.5029 -0.58107934 COORDBRK D8 (Y) - 0 - -35 COORDBRK D8 (A) - 0 - -36 STANDARD IM6 -322.2179 128.3022 MIRROR 201.0975 -65.1066237 COORDBRK D9 (Y) - 0 - -38 COORDBRK D9 (A) - 0 - -39 STANDARD IM7 423.5224 -305.8381 MIRROR 133.2641 13.7099140 COORDBRK D10 (Y) - 0 - -41 COORDBRK D10 (A) - 0 - -42 EVENASPH IM8 387.7968 397.9117 MIRROR 505.0304 0.05271243 COORDBRK D11 (Y) - 0 - -44 COORDBRK D11 (A) - 0.01073916 - -IMA STANDARD Infinity 52.63152 0
240
Table A-2. ZEMAX Coordinate Break Summary
Surface X decenter Y decenter α β γ(mm) (mm) (mm) (◦) (◦) (◦)
8 0 0 38 0 010 0 0 38 0 011 0 0 -38 0 013 0 0 -38 0 014 0 283.0357 0 0 015 0 0 1.4805 0 018 0 14.5212 0 0 019 0 0 16.906 0 020 0 -0.5 0 0 022 0 0 -4 0 024 0 0 4 0 026 0 0.5 0 0 029 0 0 0 0 18031 0 -237.0789 0 0 032 0 0 10.0149 0 034 0 21.9164 0 0 035 0 0 -9.2779 0 037 0 -47.5662 0 0 038 0 0 -12.5764 0 040 0 -5.4163 0 0 041 0 0 -3.4666 0 043 0 -49.4212 0 0 044 0 0 3.8796 0 0
241
REFERENCES
Alvarez, P., Rodriguez-Espinosa, J. M., & Kabana, F. R. 2000, in Proc. SPIE, Vol. 4004,Telescope Structures, Enclosures, Controls, Assembly/Integration/Validation, andCommissioning, ed. T. A. Sebring & T. Andersen (Bellingham: SPIE), 26
Assousa, G. E., Herbst, W., & Turner, K. C. 1977, ApJ, 218, L13
Bethune, J. D. 1997, Engineering Graphics with AutoCAD Release 13 (2nd ed.; UpperSaddle River: Prentice-Hall, Inc.)
Bibby, J. L., Crowther, P. A., Furness, J. P., & Clark, J. S. 2008, MNRAS, 386, L23
Binney, J., & Tremaine, S. 1987, Galactic Dynamics (Princeton: Princeton UniversityPress)
Bonnell, I. A., & Bate, M. R. 2006, MNRAS, 370, 488
Bonnell, I. A., Clarke, C. J., Bate, M. R., & Pringle, J. E. 2001, MNRAS, 324, 573
Brandner, W., Grebel, E. K., Chu, Y.-H., & Weis, K. 1997, ApJ, 475, L45
Carraro, G., Romaniello, M., Ventura, P., & Patat, F. 2004, A&A, 418, 525
Cepa, J., Aguiar, M., Bland-Hawthorn, J., Castaneda, H. O., Cobos, F., Correa,S., Espejo, C., Fragoso, A. B., Fuentes, J., Gigante, J. V., Gonzalez, J. J.,Gonzalez-Escalera, V., Gonzalez-Serrano, I., Joven, E., Lopez, J. C., Militello, C.,Peraza, L., Perez, A., Perez, J., Rasilla, J. L., Sanchez, B., & Tejada, C. 2003, inRevista Mexicana de Astronomia y Astrofisica Conference Series, Vol. 16, Science withthe GTC, ed. J. M. Rodriguez Espinoza, F. Garzon Lopez, & V. Melo Martin (MexicoCity: UNAM), 13
Chaisson, E., & McMillan, S. 1996, Astronomy Today (2nd ed.; Upper Saddle River :Prentice Hall, 1996)
Clark, J. S., & Negueruela, I. 2002, A&A, 396, L25
Clark, J. S., & Negueruela, I. 2004, A&A, 413, L15
Clark, J. S., Negueruela, I., Crowther, P. A., & Goodwin, S. P. 2005, A&A, 434, 949
Corbel, S., Chapuis, C., Dame, T. M., & Durouchoux, P. 1999, ApJ, 526, L29
Corbel, S., & Eikenberry, S. S. 2004, A&A, 419, 191
Corbel, S., Wallyn, P., Dame, T. M., Durouchoux, P., Mahoney, W. A., Vilhu, O., &Grindlay, J. E. 1997, ApJ, 478, 624
Cox, A. N. 2000, Allen’s Astrophysical Quantities (4th ed.; New York: AIP Press;Springer)
242
Crowther, P. A. 2007, ARA&A, 45, 177
Crowther, P. A., Hadfield, L. J., Clark, J. S., Negueruela, I., & Vacca, W. D. 2006,MNRAS, 372, 1407
Cushing, M. C., Vacca, W. D., & Rayner, J. T. 2004, PASP, 116, 362
DeGioia-Eastwood, K., Throop, H., Walker, G., & Cudworth, K. M. 2001, ApJ, 549, 578
Devaney, N., Bello, D., Femenia, B., Castro, J., Villegas Lopez, A., Reyes, M., &Fuensalida, J. J. 2004, in Proc. SPIE, Vol. 5490, Advancements in Adaptive Optics., ed.D. Bonaccini Calia, B. L. Ellerbroek, & R. Ragazzoni (Bellingham: SPIE), 913
Duncan, R. C., & Thompson, C. 1992, ApJ, 392, L9
Edwards, M. L., Eikenberry, S. S., Marin-Franch, A., Charcos-Llorens, M., Rodgers, M.,Julian, J., Raines, N., & Packham, C. 2006, in Proc. SPIE, Vol. 6269, Ground-basedand Airborne Instrumentation for Astronomy, ed. I. S. McLean & M. Iye (Bellingham:SPIE), 157
Edwards, M. L., Marin-Franch, A., Eikenberry, S. S., Rodgers, M., Julian, J., Hanna,K., & Packham, C. 2004, in Proc. SPIE, Vol. 5492, Ground-based Instrumentation forAstronomy, ed. A. F. M. Moorwood & M. Iye (Bellingham: SPIE), 1710
Eikenberry, S., Andersen, D., Guzman, R., Bally, J., Cuevas, S., Fletcher, M., Gardhouse,R., Gavel, D., Gonzalez, A., Gruel, N., Hamann, F., Hamner, S., Julian, R., Julian, J.,Koo, D., Lada, E., Leckie, B., Lopez, J. A., Pello, R., Perez, J., Rambold, W., Roman,C., Sarajedini, A., Tan, J., Venn, K., Veran, J.-P., & Ziegert, J. 2006a, in Proc. SPIE,Vol. 6269, Ground-based and Airborne Instrumentation for Astronomy., ed. I. S.McLean & M. Iye (Bellingham: SPIE), 188
Eikenberry, S., Elston, R., Raines, S. N., J., J., Hanna, K., Hon, D., Julian, R.,Bandyopadhyay, R., Bennett, J. G., Bessoff, A., Branch, M., Corley, R., Eriksen, J.-D.,Frommeyer, S., Gonzalez, A., Herlevich, M., Marin-Franch, A., Marti, J., Murphey,C., Rashkin, D., Warner, C., Leckie, B., Gardhouse, W. R., Fletcher, M., Dunn, J.,Wooff, R., & Hardy, T. 2006b, in Proc. SPIE, Vol. 6269, Ground-based and AirborneInstrumentation for Astronomy., ed. I. S. McLean & M. Iye (Bellingham: SPIE), 39
Eikenberry, S. S., Elston, R., Guzman, R., Julian, J., Raines, S. N., Gruel, N., Boreman,G., Glenn, P. E., Hull-Allen, C. G., Hoffman, J., Rodgers, M., Thompson, K., Flint,S., Comstock, L., & Myrick, B. 2004a, in Proc. SPIE, Vol. 5492, Ground-basedInstrumentation for Astronomy, ed. A. F. M. Moorwood & M. Iye (Bellingham: SPIE),1264
Eikenberry, S. S., Garske, M. A., Hu, D., Jackson, M. A., Patel, S. G., Barry, D. J.,Colonno, M. R., & Houck, J. R. 2001, ApJ, 563, L133
243
Eikenberry, S. S., Matthews, K., LaVine, J. L., Garske, M. A., Hu, D., Jackson, M. A.,Patel, S. G., Barry, D. J., Colonno, M. R., Houck, J. R., Wilson, J. C., Corbel, S., &Smith, J. D. 2004b, ApJ, 616, 506
Elston, R., Raines, S. N., Hanna, K. T., Hon, D. B., Julian, J., Horrobin, M., Harmer,C. F. W., & Epps, H. W. 2003, in Proc. SPIE, Vol. 4841, Instrument Design andPerformance for Optical/Infrared Ground-based Telescopes., ed. M. Iye & A. F. M.Moorwood (Bellingham: SPIE), 1611
Fahlman, G. G., & Gregory, P. C. 1981, Nature, 293, 202
Figer, D. F., McLean, I. S., & Morris, M. 1995, ApJ, 447, L29
Figer, D. F., McLean, I. S., & Morris, M. 1999, ApJ, 514, 202
Figer, D. F., McLean, I. S., & Najarro, F. 1997, ApJ, 486, 420
Figer, D. F., Najarro, F., Geballe, T. R., Blum, R. D., & Kudritzki, R. P. 2005, ApJ, 622,L49
Figer, D. F., Najarro, F., & Kudritzki, R. P. 2004, ApJ, 610, L109
Figer, D. F., Najarro, F., Morris, M., McLean, I. S., Geballe, T. R., Ghez, A. M., &Langer, N. 1998, ApJ, 506, 384
Fruchter, A. S., Levan, A. J., Strolger, L., Vreeswijk, P. M., Thorsett, S. E., Bersier,D., Burud, I., Castro Ceron, J. M., Castro-Tirado, A. J., Conselice, C., Dahlen, T.,Ferguson, H. C., Fynbo, J. P. U., Garnavich, P. M., Gibbons, R. A., Gorosabel, J., Gull,T. R., Hjorth, J., Holland, S. T., Kouveliotou, C., Levay, Z., Livio, M., Metzger, M. R.,Nugent, P. E., Petro, L., Pian, E., Rhoads, J. E., Riess, A. G., Sahu, K. C., Smette, A.,Tanvir, N. R., Wijers, R. A. M. J., & Woosley, S. E. 2006, Nature, 441, 463
Fryer, C. L. 1999, ApJ, 522, 413
Fryer, C. L., & Kalogera, V. 2001, ApJ, 554, 548
Fuchs, Y., Mirabel, F., Chaty, S., Claret, A., Cesarsky, C. J., & Cesarsky, D. A. 1999,A&A, 350, 891
Garzon, F., Fuentes, J., Manescau, A., Dıaz, J. J., Patron, J., Pello, R., Lopez, J. C.,Perez, J., Fragoso, A. B., Gago, F., Beigbeder, F., Sanchez, V., Correa, S., & Villegas,A. 2003, in Revista Mexicana de Astronomia y Astrofisica Conference Series, Vol. 16,Science with the GTC, ed. J. M. Rodriguez Espinoza, F. Garzon Lopez, & V. MeloMartin (Mexico City: UNAM), 23
Gavriil, F. P., Kaspi, V. M., & Woods, P. M. 2002, Nature, 419, 142
Gillespie, C. M., Jr. 1981, in Proc. SPIE, Vol. 265, Shuttle pointing of electro-opticalexperiments; Proceedings of the Seminar, ed. W. Jerkovsky (Bellingham: SPIE), 1
244
Glass, I. S. 1999, Handbook of Infrared Astronomy (Cambridge: Cambridge UniversityPress)
Golenetskii, S. V., Ilinskii, V. N., & Mazets, E. P. 1984, Nature, 307, 41
Hanson, M. M. 2003, ApJ, 597, 957
Hanson, M. M., Conti, P. S., & Rieke, M. J. 1996, ApJS, 107, 281
Harris, T. I. 1991, in Proc. SPIE, Vol. 1354, Intl Lens Design Conf, ed. G. N. Lawrence(Bellingham: SPIE), 104
Heger, A., Fryer, C. L., Woosley, S. E., Langer, N., & Hartmann, D. H. 2003a, ApJ, 591,288
Heger, A., Woosley, S. E., Fryer, C. L., & Langer, N. 2003b, in From Twilight toHighlight: The Physics of Supernovae, ed. W. Hillebrandt & B. Leibundgut, 3
Herschel, W. 1801, Philosophical Transactions Series I, 90, 393
Hill, J. M., Angel, J. R. P., Lutz, R. D., Olbert, B. H., & Strittmatter, P. A. 1998, inProc. SPIE, Vol. 3352, Advanced Technology Optical/IR Telescopes VI, ed. L. M. Stepp(Bellingham: SPIE), 172
Hodapp, K. W., Jensen, J. B., Irwin, E. M., Yamada, H., Chung, R., Fletcher, K.,Robertson, L., Hora, J. L., Simons, D. A., Mays, W., Nolan, R., Bec, M., Merrill, M., &Fowler, A. M. 2003, PASP, 115, 1388
Hoekzema, N. M., Lamers, H. J. G. L. M., & van Genderen, A. M. 1992, A&A, 257, 118
Humphreys, R. M., & Davidson, K. 1994, PASP, 106, 1025
Hurley, K., Cline, T., Mazets, E., Barthelmy, S., Butterworth, P., Marshall, F., Palmer,D., Aptekar, R., Golenetskii, S., Il’Inskii, V., Frederiks, D., McTiernan, J., Gold, R., &Trombka, J. 1999, Nature, 397, 41
Hurley, K., Li, P., Vrba, F., Luginbuhl, C., Hartmann, D., Kouveliotou, C., Meegan, C.,Fishman, G., Kulkarni, S., Frail, D., Bowyer, S., & Lampton, M. 1996, ApJ, 463, L13
Kaplan, D. L., Kulkarni, S. R., Frail, D. A., & van Kerkwijk, M. H. 2002, ApJ, 566, 378
Kaplan, D. L., Kulkarni, S. R., van Kerkwijk, M. H., Rothschild, R. E., Lingenfelter,R. L., Marsden, D., Danner, R., & Murakami, T. 2001, ApJ, 556, 399
Kaspi, V. M., & Gavriil, F. P. 2003, ApJ, 596, L71
Kleinmann, S. G., Gillett, F. C., & Joyce, R. R. 1981, ARA&A, 19, 411
Klose, S., Henden, A. A., Geppert, U., Greiner, J., Guetter, H. H., Hartmann, D. H.,Kouveliotou, C., Luginbuhl, C. B., Stecklum, B., & Vrba, F. J. 2004, ApJ, 609, L13
245
Koch, D. G. 1978, in Proc. SPIE, Vol. 147, Computer-Aided Optical Design, ed. R. E.Fischer (Bellingham: SPIE), 71
Kouveliotou, C., Dieters, S., Strohmayer, T., van Paradijs, J., Fishman, G. J., Meegan,C. A., Hurley, K., Kommers, J., Smith, I., Frail, D., & Murakami, T. 1998a, Nature,393, 235
Kouveliotou, C., Kippen, M., Woods, P., Richardson, G., & Connaughton, V. 1998b, GRBCoordinates Network, 107, 1
Kulkarni, S. R., Matthews, K., Neugebauer, G., Reid, I. N., van Kerkwijk, M. H., &Vasisht, G. 1995, ApJ, 440, L61
Lagage, P. O., & Pantin, E. 1994, Nature, 369, 628
Laros, J. G., Fenimore, E. E., Klebesadel, R. W., Atteia, J.-L., Boer, M., Hurley, K., Niel,M., Vedrenne, G., Kane, S. R., Kouveliotou, C., Cline, T. L., Dennis, B. R., Desai,U. D., Orwig, L. E., Kuznetsov, A. V., Sunyaev, R. A., & Terekhov, O. V. 1987, ApJ,320, L111
LaVine, J. L., Eikenberry, S., & Davis, J. D. 2003, in Bulletin of the AmericanAstronomical Society, Vol. 35, Bulletin of the American Astronomical Society, 1229
Low, F. J., Rieke, G. H., & Gehrz, R. D. 2007, ARA&A, 45, 43
Matthews, K., & Soifer, B. T. 1994, Experimental Astronomy, 3, 77
Mazets, E. P., Golentskii, S. V., Ilinskii, V. N., Aptekar, R. L., & Guryan, I. A. 1979,Nature, 282, 587
McClure-Griffiths, N. M., & Gaensler, B. M. 2005, ApJ, 630, L161
Mezger, P. G. 1986, Ap&SS, 128, 111
Moffat, A. F. J., & Fitzgerald, M. P. 1977, A&A, 54, 263
Morris, M., & Serabyn, E. 1996, ARA&A, 34, 645
Mountain, M., Gillett, F. C., & Kurz, R. 1997, in Proc. SPIE, Vol. 2871, OpticalTelescopes of Today and Tomorrow., ed. A. L. Ardeberg (Bellingham: SPIE), 15
Muno, M. P., Clark, J. S., Crowther, P. A., Dougherty, S. M., de Grijs, R., Law, C.,McMillan, S. L. W., Morris, M. R., Negueruela, I., Pooley, D., Portegies Zwart, S., &Yusef-Zadeh, F. 2006, ApJ, 636, L41
Munoz-Tunon, C., Vernin, J., & Varela, A. M. 1997, A&AS, 125, 183
Murakami, T., Tanaka, Y., Kulkarni, S. R., Ogasaka, Y., Sonobe, T., Ogawara, Y., Aoki,T., & Yoshida, A. 1994, Nature, 368, 127
Negueruela, I., & Clark, J. S. 2005, A&A, 436, 541
246
Neugebauer, G., & Leighton, R. B. 1969, Two-micron sky survey. A preliminary catalogue(NASA SP, Washington: NASA,)
Okumura, S.-i., Mori, A., Nishihara, E., Watanabe, E., & Yamashita, T. 2000, ApJ, 543,799
Pasquali, A., Comeron, F., & Nota, A. 2006, A&A, 448, 589
Paumard, T., Maillard, J. P., Morris, M., & Rigaut, F. 2001, A&A, 366, 466
Phillips, A. C. 1999, The Physics of Stars, 2nd Edition (New York: Wiley)
Pilbratt, G. L. 2004, in Proc. SPIE, Vol. 5487, Optical, Infrared, and Millimeter SpaceTelescopes, ed. J. C. Mather (Bellingham: SPIE), 401
Portegies Zwart, S. F., Makino, J., McMillan, S. L. W., & Hut, P. 1999, A&A, 348, 117
Portegies Zwart, S. F., Makino, J., McMillan, S. L. W., & Hut, P. 2001, ApJ, 546, L101
Portegies Zwart, S. F., & McMillan, S. L. W. 2002, ApJ, 576, 899
Pozzo, M., Naylor, T., Jeffries, R. D., & Drew, J. E. 2003, MNRAS, 341, 805
Preibisch, T., & Zinnecker, H. 2001, in Astronomical Society of the Pacific ConferenceSeries, Vol. 243, From Darkness to Light: Origin and Evolution of Young StellarClusters, ed. T. Montmerle & P. Andre (San Francisco: ASP), 791
Rayner, J. T., Toomey, D. W., Onaka, P. M., Denault, A. J., Stahlberger, W. E., Vacca,W. D., Cushing, M. C., & Wang, S. 2003, PASP, 115, 362
Rieke, G. H. 2007, ARA&A, 45, 77
Rieke, G. H., Lebofsky, M. J., & Low, F. J. 1985, AJ, 90, 900
Roche, P. F. 2004, Advances in Space Research, 34, 583
Rodrıguez Espinosa, J. M., & Alvarez, P. 2003, in Revista Mexicana de Astronomia yAstrofisica Conference Series, Vol. 16, Science with the GTC, ed. J. M. RodriguezEspinoza, F. Garzon Lopez, & V. Melo Martin (Mexico City: UNAM), 1
Rodriguez Espinosa, J. M., & Alvarez Martin, P. 2006, Astronomical Facilities of the NextDecade, 26th meeting of the IAU, Special Session 1, 1
Rothschild, R. E., Kulkarni, S. R., & Lingenfelter, R. E. 1994, Nature, 368, 432
Sabelhaus, P. A., Campbell, D., Clampin, M., Decker, J., Greenhouse, M., Johns, A.,Menzel, M., Smith, R., & Sullivan, P. 2005, in Proc. SPIE, Vol. 5899, UV/Optical/IRSpace Telescopes: Innovative Technologies and Concepts II., ed. H. A. MacEwen(Bellingham: SPIE), 241
Schroeder, D. J. 2000, Astronomical Optics (New York: Academic Press)
247
Shull, P., Jr. 1983, ApJ, 275, 611
Skrutskie, M. F., Cutri, R. M., Stiening, R., Weinberg, M. D., Schneider, S., Carpenter,J. M., Beichman, C., Capps, R., Chester, T., Elias, J., Huchra, J., Liebert, J., Lonsdale,C., Monet, D. G., Price, S., Seitzer, P., Jarrett, T., Kirkpatrick, J. D., Gizis, J. E.,Howard, E., Evans, T., Fowler, J., Fullmer, L., Hurt, R., Light, R., Kopan, E. L.,Marsh, K. A., McCallon, H. L., Tam, R., Van Dyk, S., & Wheelock, S. 2006, AJ, 131,1163
Smith, W. J. 2005, Modern Lens Design (2nd Ed.; New York: The McGraw HillCompanies, Inc.)
Spitzer, L. 1987, Dynamical evolution of globular clusters (Princeton: Princeton UniversityPress)
Stetson, P. B. 1987, PASP, 99, 191
Stetson, P. B. 1992, in Astronomical Society of the Pacific Conference Series,Vol. 25, Astronomical Data Analysis Software and Systems I, ed. D. M. Worrall,C. Biemesderfer, & J. Barnes (San Francisco: ASP), 297
Stolte, A., Brandner, W., Brandl, B., Zinnecker, H., & Grebel, E. K. 2004, AJ, 128, 765
Telesco, C. M. 2000, in Astronomical Society of the Pacific Conference Series, Vol. 219,Disks, Planetesimals, and Planets, ed. G. Garzon, C. Eiroa, D. de Winter, & T. J.Mahoney (San Francisco: ASP), 268
Telesco, C. M., Ciardi, D., French, J., Ftaclas, C., Hanna, K. T., Hon, D. B., Hough,J. H., Julian, J., Julian, R., Kidger, M., Packham, C. C., Pina, R. K., Varosi, F., &Sellar, R. G. 2003, in Proc. SPIE, Vol. 4841, Instrument Design and Performancefor Optical/Infrared Ground-based Telescope, ed. M. Iye & A. F. M. Moorwood(Bellingham: SPIE), 913
Telesco, C. M., Pina, R. K., Hanna, K. T., Julian, J. A., Hon, D. B., & Kisko, T. M. 1998,in Proc. SPIE, Vol. 3354, Infrared Astronomical Instrumentation, ed. A. M. Fowler(Bellingham: SPIE), 534
Thompson, C., & Duncan, R. C. 1996, ApJ, 473, 322
Turner, D. G., Welch, G., Graham, M., Fairweather, D., Horsford, A., Seymour, M., &Feibelman, W. 2001, Journal of the American Association of Variable Star Observers(JAAVSO), 29, 73
van Genderen, A. M., Bijleveld, W., & van Groningen, E. 1984, A&AS, 58, 537
Vancura, O., Blair, W. P., Long, K. S., & Raymond, J. C. 1992, ApJ, 394, 158
Vasisht, G., Kulkarni, S. R., Frail, D. A., & Greiner, J. 1994, ApJ, 431, L35
248
Vrba, F. J., Henden, A. A., Luginbuhl, C. B., Guetter, H. H., Hartmann, D. H., & Klose,S. 2000, ApJ, 533, L17
Vrba, F. J., Luginbuhl, C. B., Hurley, K. C., Li, P., Kulkarni, S. R., van Kerkwijk, M. H.,Hartmann, D. H., Campusano, L. E., Graham, M. J., Clowes, R. G., Kouveliotou, C.,Probst, R., Gatley, I., Merrill, M., Joyce, R., Mendez, R., Smith, I., & Schultz, A. 1996,ApJ, 468, 225
Wachter, S., Ramirez-Ruiz, E., Dwarkadas, V. V., Kouveliotou, C., Granot, J., Patel,S. K., & Figer, D. 2008, Nature, 453, 626
Werner, M., Roellig, T. L., Low, F. J., Rieke, G., Rieke, M., Hoffmann, W. F., Young, E.,Houck, J., Fazio, G., Hora, J., Gehrz, R., Soifer, T., Helou, G., Keene, J., Eisenhardt,P., Gallagher, D., Gautier, T. N., Irace, W., Lawrence, C., Simmons, L., Wright, E. L.,Jura, M., Cruikshank, D., & Brandl, B. 2004, in Bulletin of the American AstronomicalSociety, Vol. 36, Bulletin of the American Astronomical Society, 699
Wilson, J. C., Eikenberry, S. S., Henderson, C. P., Hayward, T. L., Carson, J. C., Pirger,B., Barry, D. J., Brandl, B. R., Houck, J. R., Fitzgerald, G. J., & Stolberg, T. M. 2003,in Proc. SPIE, Vol. 4841, Instrument Design and Performance for Optical/InfraredGround-based Telescopes, ed. M. Iye & A. F. M. Moorwood (Bellingham: SPIE), 451
Woods, P. M., Kouveliotou, C., van Paradijs, J., Hurley, K., Kippen, R. M., Finger,M. H., Briggs, M. S., Dieters, S., & Fishman, G. J. 1999, ApJ, 519, L139
Woods, P. M., & Thompson, C. 2006, Compact Stellar X-ray Sources, ed. W. Lewin &M. van der Klis (Cambridge: Cambridge University Press), 547
Wszolek, B., Rudnicki, K., Masi, S., de Bernardis, P., & Salvi, A. 1989, Ap&SS, 152, 29
249
BIOGRAPHICAL SKETCH
Michelle Lynn Edwards was born on December 17, 1979 in Trenton, NJ to Walter and
Linda Edwards. A voracious reader from an early age, Michelle attended Little Friends
Day School, Mercerville Elementary School, and Crockett Middle School, where she had
an excellent childhood education. When not reading, she was often studying geography or
visiting historical sites in Trenton and Philadelphia with her dad, an avid history buff. She
also loved to sing and was in choirs from a very early age.
Michelle and her younger brother Matthew Scott, always got along well. Matt grew
to be a talented musician and is now married to Jacquelyn Edwards nee O’Rourke, and
working as both a UPS driver and self-employed music engineer.
At the end of her 6th grade year, the Edwards moved to North Hanover Township,
NJ, where Michelle would spend the next six years attending Northern Burlington County
Regional Junior/Senior High School (NBC). While in high school, Michelle was interested
in law, history, and political science. She was a top debater and one of the founding
members of the NBC Debate Team directed by Mr. Joe Coleman. She also participated in
musical and theatrical productions, mock elections, and the National Honor Society. She
was a library aid and captain of the Color Guard in the NBC Greyhound marching band.
On weekends, she worked at the local farm market and bakery, Mr. McGregor’s Garden,
making pies and baked goods.
After high school, Michelle attended Dickinson College in Carlisle, PA where she
was introduced by Drs. Priscilla Laws, Robert Boyle, and Windsor (Tony) Morgan to
Workshop Physics, an activity-based course that replaced introductory physics lectures.
After excelling in the class and showing interest in astronomy, she was invited by her
adviser Tony Morgan, to travel to Flagstaff, AZ and observe with the 31” Lowell Telescope
on Anderson Mesa. The beautiful skies, excitement of observing, and especially the
instrumentation, drew Michelle further into the field of astronomy. She declared her
physics major and astronomy minor soon after returning from the observing run. The
250
next summer she completed an REU with Drs. Stephen Levine and David Monet at the
Naval Observatory in Flagstaff, AZ.
With hard work and encouragement from the faculty in the Physics and Astronomy
Department at Dickinson, Michelle earned a position as a graduate school at the
University of Florida in Gainesville, Florida. During her senior year at Dickinson, she
was inducted into Sigma Pi Sigma and Phi Beta Kappa. She graduated from Dickinson
College, Summa Cum Laude with Departmental Honors. She was the first member of her
mother’s extended family and her father’s immediate family to earn a four year degree and
the first person since her mother to attend college.
During her tenure at Dickinson, Michelle met her husband David Edmeades, only
child of Marie and Paul Edmeades. David moved to Florida with Michelle after they both
graduated from Dickinson. Avid gourmets, the couple enjoys entertaining and cooking for
their friends in the townhouse they share with their cat, Puck. After seven years together
the couple were married on October 9th, 2006.
While in graduate school, Michelle was fortunate to TA for three years for Dr. John
Oliver, an excellent teacher, good friend and confidant. She worked on her second year
project with Dr. Ata Sarajedini, who later graciously agreed to join her thesis committee,
along with Drs. Rafael Guzman, Guido Mueller, and Charlie Telesco. Then in March
2003, she began working for Dr. Stephen Eikenberry on a summer instrumentation
project. After a few months of successful work, including passing her qualifying exam
and earning her Master’s Degree, Steve took Michelle on as a thesis student to work on
the CIRCE project. In 2006, Dr. Reba Bandyopadhyay agreed to join Steve as Michelle’s
co-adviser.
Throughout her graduate career, Michelle served as president, secretary, and
admission committee liaison for the UF Graduate Astronomy Organization (GAO).
She also organized and co-organized a variety of successful public outreach programs
throughout Alachua County. During her final two years at Florida, she earned a P.E.O.
251
Scholar Award, a monetary award given to outstanding female graduate students
throughout Canada and the US, and a University of Florida McLaughlin Dissertation
Scholarship.
In August 2008, Michelle and David will move from Florida to La Serena, Chile
where Michelle will start her new position as a Gemini Science Fellow with Gemini South
Observatory.
252